Interregional Competition in the Wood Products Industry ...
Transcript of Interregional Competition in the Wood Products Industry ...
Portland State University Portland State University
PDXScholar PDXScholar
Dissertations and Theses Dissertations and Theses
1-1-1987
Interregional Competition in the Wood Products Interregional Competition in the Wood Products
Industry: An Econometric Spatial Equilibrium Industry: An Econometric Spatial Equilibrium
Approach Approach
M. Hossein Haeri Portland State University
Follow this and additional works at: https://pdxscholar.library.pdx.edu/open_access_etds
Let us know how access to this document benefits you.
Recommended Citation Recommended Citation Haeri, M. Hossein, "Interregional Competition in the Wood Products Industry: An Econometric Spatial Equilibrium Approach" (1987). Dissertations and Theses. Paper 539. https://doi.org/10.15760/etd.539
This Dissertation is brought to you for free and open access. It has been accepted for inclusion in Dissertations and Theses by an authorized administrator of PDXScholar. Please contact us if we can make this document more accessible: [email protected].
INTERREGIONAL COMPETITION IN THE WOOD PRODUCTS INDUSTRY:
AN ECONOMETRIC SPATIAL EQUILIBRIUM APPROACH
by
M. HOSSEIN HAERI
A dissertation submitted in partial fulfillment of the requirements for the degree of
DOCTOR OF PHILOSOPHY in
URBAN STUDIES: REGIONAL SCIENCE
Portland State University
1987
TO THE OFFICE OF GRADUATE STUDIES AND RESEARCH
The members of the committee approved the dissertation
of M. Hossein Haeri presented May 21, 1987.
Anthony M. Rufolo, Chairman
APPROVED:
Public Affairs
Bernard Ross, Dean of Graduate Studies and Research
AN ABSTRACT OF THE DISSERTATION OF M. Hossein Haeri for the
Doctor of Philosophy in Urban Studies: Regional Science
preserited May 21, 1987.
Title: Interregional Competition in the Wood Products
Industry: An Econometric spatial Equilibrium
Approach.
APPROVED BY MEMBERS OF THE DISSERTATION COMMITTEE:
~eSlG: Strathman
This study presents a multiregional model of the soft
wood forest products industry in the united States, designed
to describe the dynamics of interregional competition in the
industry and to provide a means for policy experimentation
and short-term projection of regional market shares. Two
2
products (softwood lumber and plywood), five product supply
regions (including Canada), and six product
are recognized. The design of the model
demand regions
is based on a
combined top-down/bot tom-up approach and consists of three
interdependent components: 1) the aggregate product market,
2) regional product markets, and 3) regional factor markets.
Model solutions are obtained by the simultaneous
determination of national level product prices and quanti
ties and allocation of equilibrium quantities across
producing regions on the basis of their relative prices and
locational advantage.
The model is evaluated in an historical simulation
using data for 1950-84. Graphical analysis of simulated
series suggests that the model replicates short-run trends
as well as cyclical movements in aggregate demand and
regional market shares. The results indicate that the short
run impacts of relative prices and locational advantage
on regional market shares are generally small. Price respon
siveness of regional market shares for lumber appear to be
considerably lower than that of plywood, indicating greater
degrees of regional sUbstitution in the plywood market.
The forecasting application of the model is demons
trated by extrapolating the complete structure for two years
beyond the sample period. The projected trend during this
two-year period is one of increasing demand for both lumber
and plywood. Domestic producers' shares of the lumber market
are expected to remain relatively stable. The results show
3
that nearly all increases in demand for lumber in this
period will be satisfied by Canadian imports.
ACKNOWLEDGEMENTS
All credit for this work is shared in equal propor
tions with my wife Mary and my daughter Mana. Without their
affectionate support and unyielding patience, I could have
never completed t.his work.
I am indebted to professors Anthony Rufolo and James
Strathman, teachers and friends, who inspired and guided me
throughout my graduate studies at Portland State University.
They helped me overcome my ignorance in many areas of
regional economics without compromising my self confidence.
My enduring gratitude goes to Dr. Richard Haynes of
the Pacific Northwest Forest and Range Experiment station in
Portland for giving me so much of his valuable time whenever
I needed it; and for his many scholarly works that lit my
way through the deep forests of the wood products economy.
Professors Tom Potiowsky and Giles Burgess of the
Economics Department at Portland State University provided
many valuable comments on earlier drafts of this document.
I also acknowledge with gratitude the receipt of
financial assistance from the School of Urban and Public
Affairs during the course of my graduate studies at Portland
State University.
TABLE OF CONTENTS
PAGE
ACKNOWLEDGEMENTS ....••.•.............•.•.....•.•..•.. iii
LIST OF TABLES ..•...•.•....•••.•.•..••.•....•...•.... vi
LIST OF FIGURES •.......•.•.••.••.••.•••....•••.••••.. viii
CHAPTER
I INTRODUCTION.... • . . . • • • • • • . • . • . • . • • • • • • . • . . • 1
The Methodological Perspective •••..••.•.. 4
The Theoretical Framework •...•..•...•..•• 7
Product Demand Demand for Regional Products Regional Product Supply Regional Factor Markets The Complete Model
II THE INDUSTRY ••..•••.•.••••..•...•..•.......• 24
The Structure of the Industry............ 25
The Market Process •••••••••.••..••.•••.•. 27
The Spatial Dimension •••.•...••..••.••... 32
Regional Market Shares ••.•••••••••••••.•. 35
Regional Timber Resources................. 42
III THE EMPIRICAL MODEL •.••••••..•.•...•...••..• 52
The Model, Its Structure, and Components. 55
Model Estimation •••..•••.•.•••..••••••..• 58
v
CHAPTER PAGE
Model Components ••••..•.••.••..••.•.••... 68
The Aggregate Product Market Regional stumpage Markets Regional Product Markets Regional Market Shares
IV MODEL VALIDATION ............................ 89
Validation in Parameter Estimates .•••..•. 89
Validation in simulation and Forecasting. 97
Concluding Remarks ••••.•••.••.••••..•.•.• 115
LITERATURE CITED
APPENDIX A: Data Sources ............................ .
APPENDIX B: Average Hourly Earnings in Manufacturing
118
125
and Lumber & Wood Products •••••••.••••••. 127
APPENDIX C: Labor productivity in Lumber and Plywood Production ............................... 128
APPENDIX 0: Producer Price Indexes for Transportation services and Lumber & Wood Products Freight ........ 1 • • • • • • • • • • • • • • • • • • • • • • • • • 129
TABLE
I
II
LIST OF TABLES
Domestic Lumber consumption by End Use category (Percent of Total Consumption)
Lumber and Plywood Production Processes
III Regional Lumber Market Shares
PAGE
28
30
(% Total Consumption) ••••.•••.••.•..••.•..••.. 36
IV Regional Plywood Market Shares (% Total Consumption) .••••••••••.••..•••.••..• 39
V Distribution of Timberlands by Region and Ownership in 1977 (MM Acres) •••••.•....•••••.• 44
VI Annual Harvest of Softwood Timber by Region and Ownership (1950-1984) ••••••••.••....•..•.• 48
VII Volume of Softwood Sawtimber by Region and Ownership 1977 (MM BFt.) ••••••.•.••...••.•••.• 50
VIII
IX
X
Parameter Estimates of the Model .............. Description of Variables Used in the Model ••..
Estimation Results for Regional Stumpage Price Relationships ................................ .
XI Estimates of Regional Elasticities of Private
60
64
75
stumpage Supply ............................... 77
XII Interregional Transport Costs for Lumber 1977 ($ per MBFt. Deflated) •••...•.....••..••. 87
XIII Interregional Tra~sport Costs for Plywood 1977 ($ per MBFt. Deflated) •••••.•.•••..••••• 87
XIV Regional Indexes of Accessibility Lumber and Plywood ............. .,.............. 88
xv Impact Multipliers for Selected Explanatory variables: Regional Lumber output (Measured as % of Total Consumption) .•.••••.••••.•..•••..••...• 93
TABLE
XVI
vii
PAGE
Impact Nultipliers for Selected Explanatory Variables: Regional Lumber Market Shares ..•....... 95
XVII Projected Values for Major Softwood Lumber and Plywood Market Variables •...•....••...•....•.. 112
XVIII Average Hourly Earnings in Manufacturing and Lumber & Wood Products ........•...... 127
XIX Labor Productivity in Lumber and Plywood Production ................................... 128
XX Producer Price Index: Transportation Services and Lumber & Wood Products Freight ••..•.•..•. 129
FIGURE
1.
2.
3.
4.
5.
6.
7.
LIST OF FIGURES
Effects of Alternative Assumptions Regarding Elasticities of Supply ••••••..••••......•...•..
Model Flow Schematic •••••.••..•••.•...•.•.•....
Major Products and Stages in the Processing of Timber ................................ ., .... .
Product supply Regions . ....................... . Product Demand Regions II II • II II • II • II II II II II II II II II II II II II • II ••
Regional Market Shares in Lumber as Percent of Apparent consumption (1950-1984) ••••.••.•••••••
Regional Plywood Market Shares as Percent of Apparent Consumption (1950-1984) ••••••••..•...•
PAGE
16
22
26
34
34
40
40
8. Canadian Lumber Production, Consumption and
9.
10.
11.
12.
13.
14.
15.
16.
17.
18.
19.
Exports ........................................ 41
Model Structure and Its Main Interactions ••.•••
Historical Simulation Total Lumber consumption •
Historical Simulation Average Lumber Price ..•..
Historical Simulation Total Plywood Consumption
Historical Simulation PPI Softwood Plywood ..•••
Historical Simulation Stumpage Price NW II II II II II II • II
• II II II • II II II
II II II • II II II II
Historical Simulation Stumpage Price SO
Historical Simulation Stumpage Price SW
Historical Simulation Stumpage Price IL
Historical Simulation Lumber Price NW • II II II II II II II II II
Historical Simulation Lumber Price SO
57
98
98
99
99
100
100
101
101
102
102
FIGURE
20.
21-
22.
23.
24.
25.
26.
27.
28.
29.
30.
Historical Simulation Lumber Price SW
Historical Simulation Lumber Price IL •.......•.
Plywood Price Douglas Fir .•••.•..•.•.•....•....
Plywood Price Southern pine .•.••.•.•.•.••.•....
Lumber Market Share Canada •.•.•.••••.•.•.......
Lumber Market Share NW
Lumber Market Share SO
Lumber Market Share SW
Lumber Market Share IL
Plywood Market Share NW
Plywood Market Share SO
· ....................... . · ....................... . · ....................... . · ....................... . . ...................... . . ...................... .
ix
PAGE
103
103
104
104
105
105
106
106
107
107
108
31. Lumber Market Share North ••.•.•••••..••.••..••• 108
This
separated
CHAPl'ER I
INTRODUCTION
is a study
markets. It
of competition
concerns the
among spatially
development and
estimation of a multiregional model of the softwood forest
products industry in the United states, designed to
describe the dynamics of interregional competition in the
industry; and to provide a means for policy experimenta
tion and short-term projection of regional market
shares.
The study is motivated by an interest to find out,
given information on the aggregate national demand for these
products, how production would be allocated among spatially
separate producers. Two products (softwood lumber and
plywood), five product supply regions (including Canada),
and six product demand regions are recognized. The design
of the model is based on a combined top-down/bot tom-up
approach and consists of three interdependent components:
1) the aggregate product market, 2) regional product
markets, and 3) regional factor markets.
Model solutions are obtained by the simUltaneous
determination of national level product prices and quanti
ties; and allocation of equilibrium quantities across
2
producing regions on the basis of their relative prices and
locational advantage. For each producing region, "locational
advantage" is measured in terms of the region's overall
accessibility in the national market.
A thirty-five year period time series data from 1950
to 1984 is used to estimate the parameters of the model.
Following an assessment of the performance of the model, the
complete structure is then extrapolated for two years beyond
the sample period.
The first step in an empirical investigation is to
identify the perspective from which it originates, and to
define explicitly the theoretical framework in which the
results may be deciphered. The principal mode of analysis
adopted here is general equilibrium. The approach can be
best described as an eclectic one: we draw from various
areas of economic thought bordering on the neo-classical
orientation, in particular, international trade and location
theory.
The perspective here is that of the analyst interested
in the dynamics of interregional competition and short-term
forecast of regional export performance. Since the approach
adopted in this study recognizes the impacts of exogenous
demand and local supply conditions simultaneously, it
presents a suitable alternative to such naive devices as
regional "shift-share" and, its counterpart in inter-
national trade, "constant market-share analysis" and serve
3
as a versatile tool for policy simulation. 1
The organization of this thesis is in four parts. The
first chapter begins by casting the research problem in the
general theoretical framework of interregional competition.
We will then proceed to explore the required methods and
necessary ingredients for developing a multiregional
industry model.
The second chapter is intended to serve as an
introduction to the forest products industry. Here, various
economic, spatial, and institutional aspects of the industry
will be examined in some detail. Institutional fa:.:tors will
be discussed only to the extent that they contribute to
behavioral variations in regional markets. We shall make no
attempt to partake in the ongoing controversy that surrounds
the issue of public policy in this area. In this part of the
study we will also specify the spatial units of our analysis
(regions), consider their characteristics and describe the
qualitative and quantitative differences that prevail among
them in terms of resource endowment and ability to grow,
process, and market their products.
1 We do not purport to discuss these techniques at any length here or elsewhere in this document. Richardson (1978) contains an ample discussion on the theory and application of "shift-share" analysis in regional studies. See also Houston (1967) for a critical treatment of the concept. For an exposition of "constant market share" analysis in international trade and a bibliography of empirical applications of the approach the reader is referred to Leamer and Stern (1970) pp.171-82.
4
Chapter III presents a summary review of previous
attempts at modeling the timber industry and use whatever
insight that can be drawn from them in developing estimating
relationships for the empirical model. In this part of the
study, after considering relevant statistical and method-
ological issues, the complete empirical model will be
estimated. Chapter IV, the final part of the study, is
devoted entirely to the analysis of results and evaluation
of performance of the model.
THE METHODOLOGICAL PERSPECTIVE
The subject of this study bears a prima facie
resemblance to the well-known Enke-Samuelson problem of
equilibrium among spatially separated markets. It seems
therefore helpful at this preliminary stage to distinguish
between the purpose and method of the present study from
those associated with theirs.
The problem of equilibrium among spatially separated
markets was first set forth by Enke in 1951:
••• There are several (originally three) regi~ns trading a homogenous good. Each reg~on constitutes a single and distinct market ••. The regions are separated but not isolated by a transportation cost per unit whic~ is independent of volume. There are no legal restrictions to limit the actions of the profit-seeking traders in each region. For each region the functions that relate local production and local use to local prices are known, and consequently, the magnitude of the difference which will be exported or imported at each local price is also known. Given these trade functions and transportation costs, we wish to ascertain: (1) the net
price in each region; (2) the quantity of imports oT. exports for each region; (3) which region exports, imports, or does neither; (4) the volume and direction of trade between each possible pair of regions." (Enke 1951)
5
Given linear regional production and demand functions,
and no economies of scale in transportation, Enke conceived
of a solution to this problem by means of an electric
analogue. Samuelson (1952) proceeded with the above
formulation and demonst~ated how it can be cast
mathematically into a maximum problem and sol ved via the
Hitchcock-Koopmans minimum transport cost linear programming
procedure. The procedure consists of an iterative procss of
varying interregional flows towards increasing "net social
payoff" (i.e. the algebraic sum of areas under local excess
supply and demand functions minus total transportation cost
of all possible flows).2
Consequent elaborations of the approach and
development of new solution algorithms have helped extend
its applications to a wide range of analytic situations
concerning temporal, as well as spatial activity and
2 For a non-mathematical discussion of the theory and application of linear programming see Dorfman (1953). Isard (1960, PP. 413-88) contains a compreh~nsive presentation of the technique and solution procedures in mathematical programming. For a more elaborate presentation and alternative formulations of spatial allocation and distribution problems see Takayama and Judge (1971).
6
allocation problems. 3 Developments have been particularly
remarkable in the direction of relaxing the stringent
assumptions of the early formulations concerning quantities
and prices, and especially, linearity of production and
demand functions. contributions of Takayama and Judge
(1964a, 1964b, 1971) have lead to the development of opera-
tional programming models in which prices and quantities are
determined endogenously within the model.
Despite their theoretical elegance and versatility,
most variants of mathematical programming models, save the
very simple single objective linear formulations, pose
serious practical problems when applied to general
equilibrium analysis where large numbers of inputs,
commodities, and locations are involved. In fact
applications of mathematical programming models in spatial
contexts have invariably been cast in a partial equilibrium
framework in which only one sector of the economy or a
limited group of related commodities is considered while the
demand for, and prices, of other goods are determined
3 The first application of linear programming in a spatial allocation setting was made by Fox (1953) in his study of the feed-livestock market in the United states. The method has since been applied to a myriad of agricultural products (See Labys 1975), various natural resource commodities (for example Kennedy 1974), the timber industry (Holley, et al. 1975, Haynes, et al. 1978), and the equilibrium analysis of international trade (Uzawa 1958, Ginsburgh and Waelbroeck 1981).
7
exogenously. And yet, even in these situations the analysis
is often hindered by the complexity of solution schemes.
The purpose of the present study is more modest and,
at the same time, more basic than what most programming
models propose to achieve. The objective of this study,
though not completely at variance with what can be obtained
in a programming approach, differs from it in several
fundamental ways. These differences become more apparent in
the subsequent discussion where we proceed with the develop
ment of our general theoretical model. For the time being it
suffices to point out that this study intends to investigate
optimization in a more simplified supply-demand framework
which does not concern interregional "flows" as such.
THE THEORETICAL FRAMEWORK
Having defined our objective, we must now explore the
means of achieving it. Again, as with the Enke formulation,
the conceived economic environment is a competitive one in
which several producers are engaged in the production of one
or more homogeneous good(s). Production takes place in
predetermined locations (regions) which are separated from
each other and from consumers by an intervening trans
portation cost. There are no barriers to trade and local
producers can compete freely in the national market. Given
the level of national demand for a good, we like to find
out: a) how regional output shares will be determined, and
8
b) given any initial equilibrium condition, what the effect
will be on regional output levels of changes in the
aggregate demand or local production conditions. Any
econometric model capable of identifying the unique impacts
of these factors on the regional output shares, must feature
several critical characteristics. As Engle (1979a) has
pointed out --though in a slightly different context-- such
a model must produce reliable estimates of the elasticity of
demand for regional products, determine the impact of
factors on the supply schedule, and estimate the elastici
ties of factor supplies.
The foregoing conceptual framework embodies several
necessary relationships that must be explicitly specified in
the operational model. These are:
a) aggregate national demand relationship; b) regional product demand relationships that can
determine regional allocations of output on the basis of each region's unique advantage (disadvantage) in relative cost and location vis-avis others;
c) regional product supply relationships; d) regional factor-market relationships; and e) an equilibrium condition that ensures the
consistency between projections at the national level of aggregate demand and regional supplies.
It also implies certain interactions that prevail among
these relationships at the interregional, as well as,
regional-national level which are to be accounted for in the
operational model. Under product market conditions with a
sufficiently large number of agents and perfectly elastic
supplies (as the case is often assumed to be in
9
international trade analysis) we normally need not concern
ourselves with these interactions. In national economies,
however, any analysis advanced on the basis of these
assumptions is likely to fall at the first fence. For one
thing, the number of producing regions for most commodities
is often small and their outputs constitute a large enough
portion of total supply so that the assumption of perfectly
elastic demand would not hold. The proper analytic framework
is therefore a quasi-competitive one where the actions of
anyone producer will have definite repercussions for other
producers (Henderson and Quandt 1980, p.200). For another,
any a priori assumptions concerning regional supply
elasticities may well prove unwarranted.
In sections that follow, we will specify these
relationships separately, elaborate their conceptual content
and theoretical foundations and explore the nature and
direction of their interdependencies.
Product Damand
Specification of demand relationships derives directly
from the general theory of consumer behavior. In its basic
form the theory explains demand as conditions that must be
satisfied if the consumer is to get the most for his money.
Given an index of consumer satisfaction (U), a set of goods
i=l ••• n) and their corresponding prices (p. ~
i=l ••• n), and a budget constraint (Y) for the consumer; the
demand function for any xi can be determined by maximizing
10
U=f (xl' x2'···, xn) subj ect to Y=P1 Xl + P2x2 + ••• + Pnxn•
Assuming that th~ second order conditions are fulfilled, the
first order conditions, together with the Ludget constraint
can produce consumer's n demand functions:
(1.1) for j=1,2, •.. n-1
which relates the demand for a commodity Xi to its own price
the prices of other commodities and income
(Henderson and Quandt, 1980, pp.18-22).
Equation (1.1) represents the general form of the
aggregate national demand relationship in our model.
Replacing Y with a vector of appropriate exogenous variables
Z; and letting Pi equal the weighted average national price
of commodi ty i, we specify the aggregate national demand
relationship in the following form:
(1. 2)
Demand For Regional Products
Equation (1.2) represents the general formulation of
commodity demand functions in a spaceless economy. The
inclusion of a spatial dimension in the analysis will
require certain modifications in this specification in order
to account for distinctions among Xi'S not only in terms of
11
their kind, but also with respect to their places of
production and consumption.
Let us suppose there are n goods being produced in r
regions. The products of each and every region are partially
consumed in the producing region itself and the rest,
assuming the presence of some excess supply, is shipped to
other regions. 4 Any complete model of the national economy,
therefore, will have to account for rn distinct products and
rn prices. That is, for each rXi (where the first subscript
represents the place of production) there will be k
different rPik's where each rPik' i.e. the price in region k
of good i produced in region r, equals the price of that
product in the producing region plus its transfer cost to
the consuming region. In equilibrium, as Samuelson (1952, p.
287) has shown, the unit price of the good in the consuming
region cannot, however, exceed its price in the producing
region plus the per unit transportation cost between the two
regions i. e. :
(1. 3) P'k < p, + t k r ~ - r ~r r
4 In order to avoid confusion, it is helpful at this point to add a slight refinement to our terminology. Here- after, following Armington (1969), we will use the term "goods" to distinguish between commodities in terms of their kind. Alternatively, the term "product" will be used to distinguish between commodities both in terms of kind and the geographic place of production or consumption.
12
from which it follows that we can conceivably define a
demand function for each region!s output, rdi , as:
(1.4) (k=l. •• n)
where k and r respectively represent locations of
consumption and production; and the second term on the right
hand side can be construed as a measure of the producing
region's overall accessibility in the market. 5
The foregoing discussion suggests how the second set
of relationships in our model, i.e. regional allocation of
output, might be derived. The key structural relationship
here is the demand for the regional output. We can expect
the quantity demanded of a product to decrease as its local
price rises. This occurs partly due to income and
sUbstitution effects. However in product markets with
several regional sources of supply, a major effect can
result from the shift to the same product from a different
region (Engle 1979b). One simple formulation of the demand
5 In most models of location analysis it is assumed that transportation inputs are proportional to the quantities being shipped implying that there are no economies of scale in transportation. However, in our model, transport inputs will be measured in terms of average per unit transport costs between trading areas, which when available, is a more realistic measure of the friction of distance as compared to transport "rates".
13
function facing a regional product is direct estimation of
every producing region's market share in all demand regions
as a function of its own price and prices of the same good
from alternative sources of supply:
Let rdik denote demand in region k of good i produced
in region r; then for a system of r producing regions and k
consuming regions we will have:
(1. 5)
ldil = f( dil , lPil' 2Pil' 3Pil' ••• , r Pi1)
ldi2 = f( di2 , 2Pi2' lPi2' 3Pi2' ••• , r Pi2)
ldik = f( dik, lPik' 2Pik' 3Pik' • •• I rPik)
2dil = f( dil , 2Pil' lPil' 3Pil' · .. , rPil)
2di2 = f( di2 , 2Pi2' lPi2' 3Pi2' · .. , r Pi2)
which gives a set of rk simultaneous equations in rk
endogenous variables for each coromodi ty i. Estimation of
these equations, however, poses serious practical problems.
First, data on regional demands are virtually non-existent.
Although regional consumption patterns may be derived
indirectly from production and interregional shipments, the
shipment data themselves are ei ther scarce or irrevocably
14
distorted due to intermediate storage and transshipment. 6
Second, co1linearity among the price variables will most
likely reduce the efficiency of parameter estimates
considerably.
One way to circumvent these problems is to modify the
equations in (1.5) into a form that is compatible with
available data and also more agreeable to the exigencies of
estimation. A simple approach is aggregation of rdik's over
all k's and replacing the price variables by an appropriate
measure of comparative cost. Thus reducing (1.5) to only r
equations and adding transport costs from (1. 4) we will
have:
10 i f( d (Pi -1Pi) L t1k / n = Qi'
20 i f( d (Pi -2Pi) L t2k / n = Qi'
(1. 6)
rOi f( d (Pi -rPi) , L trk / n ) = Qi'
6 For a detailed discussion of problems associated with data on in~erregional flows see Garnick (1980). Adams, Haynes, et al. (1979) have estimated regional demands for lumber and plywood in six u. S. regions for 1950-76. Their method is, however, complex and difficult to replicate. Work on updating the data is presently underway at the Pacific Northwest Forest and Range Experiment station in Portland, Oregon.
15
where rDi = 2: rdik, rPi is the regional price of i as
defined in (1.4) and Pi' as before, is the weighted average
price of the same good in the national market.
Regional product demand relationships in our model
follow the same form as in (1.6), which relates the demand
for a region's output to the total demand for that product
and the region's price competitiveness vis-a-vis the
corresponding price of the same good in the national market.
Note that the coefficients of dQi,s measure the elasticity
of regional output relative to the industry output and the
price elasticities, i.e. the coefficients of the price terms
in parentheses, measure the competitive component of the
regional market shares. 7
Regional Product Supply
Regional product supply functions constitute another
essential ingredient in our model. As we have already noted,
any model of spatial competition pivots on two elements of
locational advantage and local prices. If a region lowers
7 This coefficient is conceptually analogous to the "elastici ty of import substi tution" , the measurement of which has been the subject of perennial interest --and controversy-- in the international trade literature (Tinbergen 1946, Polak 1950, Ginsburg and stern 1965). Several alternative formulations of the relationship between the two price terms have been suggested in the lierature. Two formulations that immediately come to mind are: a) to cast the relationship in ratio form, i.e., ( p /p ), or b) presentation of the two prices separately in theie~ation. These possibilities will be explored further as we proceed to develop the exact estimating equations in the following sections.
16
(raises) its price, the quantity demanded of its product
will increase (decrease) due to the fact that other regions
will sell less (more). The regional differences in price
depend largely on relative elasticities of supply. The point
is demonstrated graphically in Figure (1). The figure shows
the market& for a homogeneous product in two regions a and
b. Both markets are in equilibrium with equal market shares
(Qlto = Q2tO)· The demand schedules Dto are sloped smoothly
because due to the aforementioned "regional sUbstitution
effect" are expected to be more elastic than the industry
demand (Engle 1979a). Sa and Sb are the supply schedules in
the two regions. The only difference between the two markets
is that the slopes of supply curves are not the same.
p p
I I I I Q Q
(a) QltO (b) Q1t1 Q2t1
Figure 1. Effects of alternative assumptions regarding elasticities of supply.
Suppose that the demand schedule shifts from Dto to
Dt1 • The subsequent resul ts of the shift and its
implications in terms of output and market share clearly are
not the same for the two regions. This differential impact
is quite evident in the difference between (QltO-Qltl) and
17
(Q2tO-Q2tl) which has resulted from unequal elasticities of
supply in the two regions.
Supply functions are often much more diverse in nature
than demand functions. They are derived in a fashion
analogous to consumer utility maximization, that is,
maximization of output by the producer subj ect to certain
cost considerations (although this is not the only mode of
optimizing behavior for the producer). Supply functions, as
distinct from production functions that relate quantities of
output to various inputs, describe the response of output to
price, factor costs and other technological, institutional,
or ecological factors (Labys 1973, ch.3). In the most
general case a supply function may therefore be specified in
the following form:
(1. 7)
where Si is the quanti ty supplied of product i, p. ~
is the
price of i, w. ~
to wn represent various input prices. Note
that all variables carry regional subscripts that have been
omitted for convenience.
Regional Factor Markets
Two factors have so far been identified that determine
a region's market share in the national economy: location
and price of outputs. Since locations are predetermined,
competitiveness of a region is ultimately determined by the
18
level of productivity in its industries and their cost
performance relative to their competitors in other regions.
It is apparent from equation (1.5) above that the level of
output is in part determined by the price of inputs. If
factor supplies are perfectly elastic, then we could assume
their prices as fixed and need not worry about what goes on
in the local factor markets. However, in reality (and
consistent with economic theory), returns to a factor of
production are market determined, i.e, in equilibrium they
can be represented as the intersection of demand for and
supply of the factor.
Factor markets are therefore inseparable from the
product market. The key to a better understanding of the
linkages between the two markets is to develop a conceptual
framework in which the relationship between prices in the
two markets can be explicitly defined.
The relationship between product price and input
prices is straight forward and can be derived from
producer's profit maximizing behavior. The producer's total
revenue is determined by the quantity of output (q)
multiplied by the unit market price (p). Producer's profit
is defined as the difference between total revenue and total
cost. Thus in a production process involving two factors,
letting xl and x2 be the levels of input of each factor 1
and 2 with wl and w2 representing their respective unit
prices, we have:
19
(1. 8)
where,
(1. 9)
The producer's demand function for each input can be derived
directly from the profit ;unction by substituting for q from
(1.9) , setting the partial derivatives of the new equation
with respect to xl and x2 equal to zero, and then solving
for xl and x2 • See for example Intriligator (1971, p. 191)
and Henderson and Quandt (1980, p. 80). Thus for a given
production function, the demand for a factor may be
expressed in terms of its own price, the price of other
inputs, and price of output, i.e.:
(1.10) for i=1. •. n
and j=1. •• n-1
This relationship implies that a producer's willing
ness to pay for factors is essentially constrained by the
price he is able to obtain for his product. Validity of this
conclusion is intuitively clear and consistent with the
derived demand concept.
On the other side of the market are the suppliers of
factors who, very much like all other producers, are also
concerned with getting the most for their lot, that is,
20
maximizing their profit (or present worth as in the case of
natural resource commodities). Their behavior can therefore
be explained within the framework of conventional static
supply theory. In the case of natural resource commodities,
however, the requisite for profit maximization is that the
opportunity cost of holding reserves also be accounted for.
One way of expressing the latter in a form amenable to
empirical estimation is the use of proxies such as total
available stocks or inventory. In our analysis we suggest to
specify the factor supply relationships in the following
form:
(loll)
Where sF i is the quantity supply of factor i in a given
region, wi is the corresponding factor price, xi represents
a measure of available stock of factors or inventory, and X
is a vector of exogenous institutional, or otherwise non
economic factors.
The complete Hodel
If there are r producing regions, n commodities and m
factors, then relationships (1.2), (1.6), (1. 7), (1.10), and
(loll), constitute a system of 2rn + 2m + 1 equations in
2rn + 2m + 1 endogenous variables, namely, rn commodity
prices (Pi)' rn regional
(Wi)' m regional factor
supplies (Si)' m factor prices
supplies (SFi ) , and 0i. For
21
convenience and ease of reference the complete system of
equations developed so far is rewritten belotor :
Aggregate Demand: dQ . l. = f( Pi' p. ,
J Z )
Demand for Regional Products: D. f( d (Pi-rPi)' = Q. , r l. l.
I trk In )
Regional Product Supplies: rSi = f( Pi' wI·· .wn '
Regional Mkt. Equilibrium: rDi = rSi
Regional Factor Demands: dF~ = f( wi' wj , p. ) .... l.
Regional Factor Supplies: sF. = f( wi' Xi' X ) l.
The major elements and interactions of the model are
given in Figure 2, which shows a schematic outline of the
model structure. In addition to the behavioral relationships
discussed above, the influence of imports on the product
market and the impact of institutional factors on regional
factor markets are also recognized.
The model developed here is intended to serve as a
general framework for the analysis of interregional
competition. In the jargon of econometric model building our
approach here is commonly described as top-down modelling,
as distinguished from the bottom-up approach (see Ballard,
Glickman, and Gustely 1980; also Milne, Adams, and Glickman
1980). The model relies on the national economy to determine
the overall level of sectoral activity which is then
allocated across producing regions on the basis of their
relative costs and locational characteristics. Our proposed
22
formulation, however, departs from a purely top-down
structure in that it allows for feed-backs from regional
economies through price variables.
Exogenous .. Aggregate Variables Prod. Demand
t Imports ~ NatIonal Mkt. .... po Price
+ It , .. ~egional Prod. .... ... !Regional Prod. ... Demand for
Supply .... .. Prices ... .. lRegional output
+ 1
Regl.onal Regional ~ ~ Regional Factor Demand .. Factor Prices .... .. Factor Supply
I ·t IInstitutl.onal Inventory Constraints
Figure 2. Model flow schematic
Before we proceed with the application of this model
to actual data, the commodity dimension of the model must be
fully developed. Therefore, this model can only serve as a
prototype in the most general sense of the term. Further
more, all relationships have been presented in their static
forms. From an empirical point of view, no conclusions can
be reached with regards to the appropriate functional forms
or adjustment processes before the nature of the commodity
under investigation is fully understood. A thorough
23
discussion of these issues will be undertaken in the
following sections where we will apply the model to the
analysis of interregional competition in the wood products
industry in the united states.
CHAPTER II
THE INDUSTRY
The choice of the softwood forest products industry
for the present study was motivated by the industry's
several characteristics that render it suitable for the
analysis of interregional competition. First, it is an
important industry. In 1984, the lumber and wood products
industry accounted for nearly four percent of total employ
ment, and three percent of value added in manufacturing in
the United states. The industry's contributions in terms of
employment and income are particularly pronounced at the
local and regional levels. In states such as Oregon and
Washington, for example, respectively forty and close to
twenty percent of manufacturing employment originates in the
timber related activities. 1
Second, it is a highly competitive industry character
ized by a very large number of firms with relatively low
1 The Pacific Northwest states rank highest in their dependence on the forest products industries. In other major producing regions such as the South, local economies are much more diversified. Among the Southern states, Arkansas, where the forest products industries account for slightly over sixteen percent of basic employment, is the most timber dependent state in the region (Scha11au and Maki, 1986).
25
levels of concentration. In 1984, for example, there were
over thirty thousand firms in the industry nationwide.
Third, the industry is also highly localized:
roundwood is heavy, bulky, and the manufacturing processes
for most products entail considerable losses in weight, as
well as, in volume. The manufacturing facilities are,
therefore, invariably located near the sources of timber
supply concentrated in few distinct regions in the country.
The forth and final factor in selecting this industry is
product homogeneity; a condition (assumption) that is
central to the analysis of interregional competition.
certain differences in grades and species notwithstanding,
the main products of the industry such as lumber, plywood,
and pulp, are sufficiently similar with respect to specific
end uses to be considered homogenous.
THE STRUCTURE OF THE INDUSTRY
The softwood forest products industry is a complex and
multi-faceted system that encompasses a wide range of
activities extending from forest lands to the consumption of
final products. This system is shaped and regulated by the
confluence of many economic, biological, institutional, and
spatial factors.
Broadly defined, the industry consists of wide a range
of activities in which timber, in combination with other
inputs (capital, labor, energy, etc.), is used to produce a
myriad of consumer products. The most commonly accepted
26
classification of these products is in terms of the stage of
processing, and involves two categories of primary and
secondary products. The main processing stages and maj or
primary products of the forest products industry are shown
in Figure 3.
Growing stock
I .. Roundwood
1 . + Ml.sc.Products
+
+ Fuelwood ~
.. Residue
+ Sawlogs Veneerlog Pulpwood
I Woodpulp ..-__ ----.1.
+ Lumber Plywood Paper
Figure 3. Major products and stages in processing of timber.
Roundwood products, which are used primarily in the
production of lumber, pulp, and plywood, claim the largest
portion of annual removals of timber in the united states.
Forest Service estimates show that during the 1970's, nearly
95 percent of annual harvest of softwood timber was in the
form of roundwood. Miscellaneous products such as poles,
pilings, posts, mine timber, etc., accounted for an
additional 2.5 percent; and the remaining 2.5 percent was
divided in near equal proportions between fuelwood and
residual products. (USDA/Forest Service, 1982).
27
The present study focuses on two of the primary forest
products: softwood lumber and plywood. These products are
two of the most important outputs of the forest products
industry in the U. S. and together account for about 70
percent of annual consumption of softwood timber in the
country.
THE MARKET PROCESS
In a free market, "industrial competition" is defined
in part as a condition wherein every productive resource in
an industry "earns as much, but no more than, it would in
other industries" (Stigler, 1968, p. 10). This condition
holds because the profit maximizing behavior of owners
guarantees that resources are employed where they obtain the
highest returns. Timber is a versatile raw material which is
used in the production of a wide range of products. At any
given time, the producers' economic considerations, 1. e.
profits, determine the type of product into which timber is
likely to be transformed.
The level of output for each product is in turn
determined by market conditions. Producers supply what
consumers
mechanism.
quantities
demand of them. Prices are the equilibrating
For each product, market clearing prices and
are dictated jointly by consumer demand and
supply conditions.
Demand: Softwood lumber and plywood are producer
goods. Demand for these products is, therefore, a derived
28
demand and depends largely on the level of activity in their
respective end use markets. The principal end uses for
softwood lumber and plywood and levels of demand in each
market in selected years are shown in TABLE I.
TABLE I
DOMESTIC LUMBER CONSUMPTION BY END USE CATEGORY (PERCENT OF TOTAL CONSUMPTION)
Lumber 19521 19844
Residential construction ••• : 36.7
Non-residential construction ••• : 18.7
Repair and alterations •••• : 16.5
Material handling and shipping ••• : 16.0
Manufacturing and other •••••• : 12.0
Plywood
Construction •••• : Manufacturing
and other •••••• :
51.0
49.0
45.4
9.90
15.2
14.1
12.8
55.4
44.6
34.1 38.1 38.8
10.2 10.5 15.4
13.0 13.3 26.3
15.8 16.1 10.0
12.9 10.1 9.50
59.1 65.7
40.9 34.3 31.0
1- SRI (1954); 2- USDA/Forest Service (1973); 3- USDA/ Forest Service (1982); 4- wes~ern Wood Products Assoc. (1984); 5- Am. Plywood Assoc.
2 It is important to note that the data on consumption patterns are based on estimates derived from different sources. In several instances comparison of data across various sources show considerable discrepencies. For example in a 1982 survey of lumber use in non-residential construction (Spelter and Anderson, 1985) lumber consumption in this market is estimated at around seven percent of total domestic consumption.
29
Historically, the construction industry has been the
single most important source of demand for softwood lumber
and plywood. As figures in Table I indicate, since 1950, the
construction sector has accounted for an average of 70
percent of all softwood lumber and 60 percent of all
softwood plywood consumed in the united states. Witi.lin the
construction industry, new residential construction is
responsible for the largest portion of lumber and plywood
consumed in this sector.
other major uses of lumber and plywood are in shipping
and manufacturing (mostly wood furniture). During the past
three decades, the use of lumber and plywood in these
activities has followed a continuously declining trend. The
declining role of timber products in activities other than
construction is particularly evident in the historical
pattern of plywood consumption.
Supply: Levels of output in a competitive industry are
determined by the producers' ability to respond to changing
demand and factor supply conditions. In order to succeed in
a competitive environment, the industry has to continuously
improve its productivity so as to decrease its production
costs subject to technological constraints inherent in its
production process.
The production of lumber and plywood is each comprised
of a series of operations in which roundwood is converted
into the final product. These operations are summarized in
Table II. During the time period covered in the present
30
analysis, most of these operations have undergone consider
able technological change.
TABLE II
LL~BER AND PLYWOOD PRODUCTION PROCESSES
Lumber
1) Log handling 2) Breakdown 3) pipping & Trimming 4) Crosscutting 5) Drying 6) Surfacing
Plywood
Log handling Peeling or Slicing Drying Assembly lay-up Glueing Finishing
Technological improvements in the wood products
industry have been char~cterized by Heady (1952) as being
either "biological", or "mechanical". According to this
classification biological innovations are essentially of the
"neutral" type and tend to increase the marginal output of
all inputs without altering their sUbstitution properties.
Improvements in plant lay-out, enhanced waste utilization
techniques, and the use of thinner saw kerfs are all but few
examples of this type of technical progress. Mechanical
innovations, on the other hand, are resource saving changes
that raise the marginal product of specific inputs relative
to others. Some examples of the latter type of technical
change are mechanized log handling techniques, computerized
log scanning methods, and improvements in small-log peeling
technology in veneer production.
31
The structure of production in the lumber and wood
products industry has been analyzed in several recent
studies including those by Robinson (1975), Humphrey and
Moroney (1975), Alperovich (1980), stier (1980), Merrifield
and Haynes (1983, 1984), and Nautiyal and Singh (1985). With
the exception of Alperovich (1980), and Merrifield and
Haynes studies of the Pacific Northwest forest products
industry, these studies generally involve analysis of broad
product categories at the aggregate national level. There is
therefore little information on patterns of technical change
associated with specific products in various regions.
since these studies have employed different production
functions and involve different product categories, their
results are not always directly comparable. They do,
however, reach some general, though important, conclusions
that bear directly on our analysis. The first conclusion is
that the rate of technical progress and productivity gains
in the lumber and wood products industry has at least
paralelled those of other manufacturing industries in the
country. Second, due to rapid increases in labor and
stumpage costs, the trend in technological change has tended
towards labor --and to a lesser extent stumpage-- augmenting
improvements. Third, most factor inputs in the industry are
substitutable; and, that sUbstitution possibilities between
capi tal and labor, and capital and stumpage, are greater
than those of other inputs.
32
with respect to scale effects, however, the results of
these studies remain by and large inconclusive. Merrifield
and Haynes (1984), for instance, find no evidence of scale
effects in the industry; while Robinson (1975) concludes
that gains in productivity have been largely attributable to
economies of scale.
THE SPATIAL DIMENSION
The market processes disscussed in the preceeding
section determine the quantity and type of product that is
likely to be produced at any given point in time. In an
industry characterized by a geographically fragmented
market, the very same forces also decide "where" each
product will be produced.
In the softwood products economy, this locational
dimension arises because of the spatial imbalance of supply
and demand resulting from an uneven distribution of
producers and consumers of forest products throughout the
united states. Historically, the consumers of forest
products have been concentrated in the Northern states,
while major sources of timber supply and production
facilities are located in the South and the west.
The two sets of supply and demand areas, together with
the transportation costs that connect them, constitute a
dynamic system. The supply regions compete with one another
to market their products in supply areas. Prices are again
the equilibrating mechanisms. If demand for a region's
33
product is excessive relative to production possibilities,
the price in that region will rise leading demand to shift
to other producing regions. This will in turn lead prices in
the initial region to fall until a new equilibrium is
reached.
In the present study, this spatial dimension is
represented by partitioning the continental united states
into a set of contiguous regions. Five domestic lumber
supply regions (Pacific Northwest, South, Southwest,
Inland, and North), two plywood supply regions (NW and SO),
and six product demand regions are considered. See Figures 4
and 5. Lumber output from the North region is treated as
exogenous. Product supplies from Alaska and Hawaii are not
included because their contributions to the domestic market
are small; and regional exports of plywood are ignored.
The main factor underlying the development of this
spatial framework' was the practical considerations of data
availability. A basic consideration in specification of
supply regions was that activities and product types should
possess some degree of homogeneity. The supply region
bounda~ies specified here correspond closely with the Forest
service Administrative Regions and are essentially identical
to supply areas considered in the Timber Assessment Market
Model (Adams and Haynes 1980). The choice of demand regions
was determined entirely by the availability of transport
costs data.
34
Figure 4. Product supply regions.
Figure 5. Product demand regions.
35
REGIONAL MARKET SHARES
Market shares are here defined as the percentage of
total domestic demand supplied by a producing region. Total
demand on producers in a supply region (regional production)
is determined by combining market shares wi th exports to
other countries.
Tables II and III present market share data covering
the 1950-1984 period for softwood lumber and plywood
produced in various supply regions. Examination of the these
historical trends in market shares reveals that significant
changes in distribution of production across various regions
has taken place.
Pacific Northwest: The main lumber products of this
region come from Douglas Fir which is especially prized for
its structural properties. Until 1965, the Pacific Northwest
enjoyed a relatively stable position in the domestic
softwood lumber market, producing annually about one third
of total softwood lumber consumed in the country. Since
then, save minor increases in 1974 and 1975, the region's
market share has declined steadily (see Figure 6).
Data on shipments of lumber from the Northwest show
that the region's loss in market share in recent years can
be primarily attributed to losses in the Northeast and
Northcentral markets. Rapid rises in transport costs
combined with increases in stumpage prices in the Northwest
have placed the region in great disadvantage vis-a-vis other
36
TABLE III
REGIONAL LUMBER MARKET SHARES (% OF TOTAL CONSUMPTION)
Market Shares Consumption
(MMMBFt. ) NW SO SW IL NO CAN Year 1950 33.4 34.7 30.6 12.6 7.3 5.5 8.7 1951 30.9 36.4 28.1 15.4 7.0 5.9 6.7 1952 31.9 38.5 27.9 14.0 7.7 4.9 6.7 1953 31.6 37.5 25.1 15.9 8.2 5.4 7.6 1954 31.6 36.5 24.7 16.0 9.2 4.7 8.7 1955 32.5 36.7 24.0 16.1 9.2 4.0 9.9 1956 32.7 32.6 25.4 17.7 10.7 4.1 9.4 1957 29.2 33.3 23.5 18.1 10.3 5.7 9.1 1958 30.0 35.7 21.5 17.4 10.9 4.1 10.3 1959 33.6 34.6 21.3 17.7 11.5 3.9 10.9 1960 29.6 36.0 19.8 17.0 11.0 4.0 12.1 1961 29.4 34.8 19.6 16.8 11. 3 4.0 13.4 1962 30.7 35.1 19.2 16.0 11.4 3.7 14.7 1963 31.7 34.4 19.3 15.5 11.4 3.9 15.7 1964 33.3 35.4 19.6 15.3 12.0 3.2 14.6 1965 33.3 35.2 20.2 14.8 12.2 3.1 14.6 1966 32.6 34.4 20.4 15.0 12.5 3.3 14.5 1967 31.8 33.9 20.7 14.9 13.0 3.2 14.9 1968 34.8 33.0 20.0 15.2 12.8 3.2 16.5 1969 32.9 29.2 22.2 15.4 13.3 3.2 17.6 1970 31.6 29.1 21.8 15.8 12.9 3.2 18.1 1971 36.0 29.0 21.8 14.8 12.3 2.8 19.9 1972 38.4 28.1 20.8 14.6 11.4 2.6 23.1 1973 38.4 28.2 20.6 14.6 11.6 2.9 23.0 1974 33.2 31.2 21.0 13.6 11.4 3.3 20.3 1975 30.2 29.3 23.1 13.5 12.7 3.6 18.8 1976 36.4 28.0 21.7 13.0 12.3 4.3 21.8 1977 40.8 26.4 21.6 12.1 11.4 4.1 25.3 1978 42.8 25.4 21.6 11.1 10.4 4.4 27.5 1979 41.0 24.9 22.7 11.0 10.0 4.6 27.1 1980 34.5 23.2 23.8 10.6 9.9 5.1 27.7 1981 32.1 23.4 26.4 9.7 9.7 3.5 28.7 1982 31.4 22.1 27.9 9.2 8.8 3.4 29.0 1983 41.5 22.9 24.6 8.9 10.2 3.1 28.8 1984 43.0 23.4 24.7 8.9 9.4 3.0 30.8
Source: Adams, Jackson, and Haynes (1986); Western Wood Products Association statistical Yearbook.
37
producing regions, leading to significant encroachments on
this region's share of the Northern market.
Recent trends in lumber production in the Northwest
do, however, suggest that the region's market shares during
the 1980's have been somewhat stable, fluctuating around a
historical low level of about 23 percent.
The South: The historical trend in lumber production
and market shares for the South are markedly different from
what we have observed in the Northwest. During the decade of
50's, lumber production in the South declined steadily
resulting, on the average, in an annual loss of about one
percent in market shares. After a relatively long period of
stability in the sixties and the greater part of the
seventies, however, the region's market shares began to rise
considerably. In 1983 lumber production in the South reached
an historically high level of 10.5 billion BFt. and rose to
near 11 billion BFt. in 1984.
The Southern region covers a significantly large
territory and, unlike the Northwest, the largest segment of
its market consists of intra-regional shipments. The most
significant change in market shares of the Southern lumber
in recent years was associated with the region's performance
in 1981 and 1982. During these years, the market share of
Southern producers increased by nearly 2.5 points per year.
The performance of the South in the plywood market has
been much more astounding. During the past three decades,
plywood consumption in the Uni ted States has grown
38
consistently at an average rate of near 500 million square
feet per year. Prior to 1964 virtually no plywood was
produced in the South. In this period, the Northwest
producers, located mostly in the Coastal region, comfortably
commanded near 90 percent of the domestic plywood market.
Refinements in small-log peeling technology during the
1960's, induced by the rapid growth in plywood demand,
provided the grounds for the development, and subsequent
expansion, of the industry in the South. Production figures
for the Southern plywood show that since 1965 the South has
not only absorbed nearly all increases in domestic demand
for softwood plywood, but indeed has encroached significant
lyon the Northwest's market shares.
Market share figures for the two regions show that the
Northwest entered the 1970's with a market share of almost
three times that of the South (see Figure 7). Currently,
South's market share stands 11 points above the Northwest.
Southwest: Softwood lumber production in the SW has
remained at a more or less steady level of approximately 4.5
billion BFt. a year. During the 1950's and 1960's, the
species composition of lumber produced in the region
--consisting of large quantities of redwood-- allowed it to
maintain a relatively stable share in the domestic market.
In recent years, however, decl ines in redwood production
have made it more and more difficult for the region to keep
pace wi th increases in domestic demand, leading to
sUbstantial losses in market shares.
39
TABLE IV
REGIONAL PLYWOOD MARKET SHARES (% OF TOTAL CONSUMPTION)
Market Shares Consumption (MMSq. Ft.) NW SO SW IL
Year 1950 2554.00 93.10 0.00 6.90 0.00 1951 2867.10 91.60 0.00 8.40 0.00 1952 3050.00 88.70 0.00 11.30 0.00 1953 3671. 00 88.20 0.00 11.80 0.00 1954 3903.60 87.40 0.00 12.60 0.00 1955 5075.00 87.40 0.00 12.60 0.00 1956 5239.20 86.80 0.00 13.00 0.30 1957 5459.50 86.90 0.00 12.90 0.20 1958 6340.30 87.20 0.00 12.60 0.20 1959 7827.80 85.40 0.00 14.00 0.60 1960 7815.60 84.40 0.00 14.60 1. 00 1961 8576.60 84.20 0.00 13.70 2.10 1962 9513.10 84.80 0.00 12.70 2.50 1963 10216.00 84.30 0.00 12.20 3.50 1964 11678.40 83.90 0.70 11.20 4.20 1965 12447.00 81.00 3.20 9.90 5.90 1966 13056.00 75.70 8.70 8.70 6.80 1967 12958.40 71.30 13.50 6.80 8.20 1968 14694.60 70.50 15.80 6.10 7.20 1969 13694.00 66.00 20.60 6.30 6.70 1970 14340.00 64.80 22.80 5.80 6.20 1971 16634.90 61.40 26.10 5.90 6.10 1972 18324.00 59.40 28.70 5.70 5.80 1973 18305.20 58.70 30.10 5.30 5.60 1974 15877.00 56.10 32.00 5.30 6.30 1975 16050.00 53.90 35.00 4.00 6.70 1976 18440.00 53.20 36.70 3.30 6.60 1977 19376.00 52.20 38.10 2.80 6.50 1978 19964.00 51.60 39.30 2.60 6.20 1979 19653.00 49.10 42.10 2.40 6.10 1980 16311.60 46.10 45.00 2.00 6.70 1981 16731.40 41.50 49.30 2.10 6.80 1982 15805.40 39.70 53.30 1.20 5.50 1983 19525.60 41. 70 50.70 1.30 5.90 1984 19901. 80 41.80 52.00 1. 30 4.40
Source: Adams, Jackson, and Haynes (1986).
40
100,----------------- --------------------------~
90 Canada
so ~North
70
Inland so
SO'Jthwest 50
South
JO
20
Northwest 10
Ig5019521954195819581g8019.2'8&41ge81868'8701872'9741978,9781980,9821984
Figure 6. Regional market shares in lumber as percent of apparent consumption (1950-84). Source: Table III.
100 I n I clnd
Southwest 90
eo
70
so J I
South
50
40
30
Northwest 20
10
0 '.50,852'854,958,958,980'982'184,181,111,.70,.72,,74,978,878,980198219&4
Figure 7. Regional Plywood market shares as percent of apparent consumption (1950-84). Source: Table IV.
41
Inland: Most lumber produced in this region comes
from the Northern Rocky Mountains part of the region. Minor
fluctuations notwithstanding, Inland's share of the domestic
lumber market has remained relatively flat since the early
1950's. Lumber production figures for the Inland region
suggest that changes in regional output correspond somewhat
closely with fluctuations in demand for lumber at the
national level leaving region's market shares unaffected.
Canadian Imports: The United states is the largest
single market for the Canadian softwood lumber (Figure 8).
Indeed Canadian exports to th~ U.s. constitute the largest
flow in international lumber trade (Lindell, 1979).
Imports of lumber from Canada have expanded
continuously since 1950. This expansion has brought about
21 20 111 111 17 HI 15 14 13 12 II 10 II II 7
II
5 4
:5 2 1
o v
'J p: 1\ f\
f\
v v
i% ~~ ~ 1\ 1\ f\ 1\ r, . ,
v V ~I V v VVv
.~
17: ~ ~ ~
~ l% I%~ f\
1\ I'
~" t
U' 'i ~ ~ ~ r, 1\
1\ h .1\ r-
v v
V l/ . V 'V
' .. 'V . ~' . V ~/ V V v'V
11I50III~2111!54111!5SIIl!5llIIlS01I1S211154111651115111111011112111141111511111111111018112111114
IZZl C<lnlumptlon lSSI Exp. 10 u.s, ~ Other Exp.
Figure 8. Canadian lumber production, consumption, and exports. Source: Adams, Jackson, and Haynes (1986).
42
equal, but opposite, changes in market shares of domestic
producers particularly in the West. Traditionally, imports
from Canada have served a market clearing function
fluctuating with changes in demand in the domestic market.
In recent years, however, these imports have exhibited a
behavior suggestive of an increasingly competitive
displacement of domestic production.
The increase in Canadian imports in recent years and
the strengthening of i ts competitive position are often
attributed to declines in the value of the Canadian dollar
in relation to u.s. currency. However, as Adams and Haynes
(l980a) have demonstrated, an equally important factor has
been the production conditions in Canada characterized by
low production costs and elastic supply.
REGIONAL TIMBER RESOURCES
A fundamental consideration in the analysis of the
spatial patterns of production in the wood products industry
is the geographical distribution of timber resources and
variations in supply behavior that arise from differences in
economic objectives and management practices across
ownership categories.
The continental united states contains about 470
million acres of timberland with an additional 12 million
acres in Alaska and Hawaii, which together constitute around
22 per cent of the total land area in the country.
43
Timberlands, as listed in Table V, are categorized by
region and type of ownership. The rows in Table V show the
distribution of timberlands within each region by ownership
type. Four types of ownership are recognized: "National
Forests" , "other public", "forest industry", and "other
private". "National Forests" designates commercial
forestlands that are owned by the Federal Government and
come under the jurisdiction of the Forest Service
(Department of Agriculture). National forests are chiefly
concentrated in the western regions where they consti tute
nearly 53 percent of all timberlands in the region.
The "other public" category includes Federal lands
other than National Forests (primarily lands administered by
the Bureau of Land Management, Department of the Interior),
Bureau of Indian Affairs; and state, county, and municipal
forests. Much of the latter holdings (over 50 percent) are
in the Lake States and consist of lands reverted to local
governments as a result of tax delinquencies during the
30 I s. The BLM lands are primarily located in the Pacific
Northwest (4.1 MM acres) and the Rocky Mountain (1.7 MM
acres) regions.
The "forest industry group", Le., integrated firms
that engage in both growing and processing of timber, holds
approximately 15 percent of total timberlands. The remaining
59 percent (277 MM acres) of timberlands in the continental
united states are owned by farmer and other private
44
entrepreneurs who grow and sell timber but in general~. are
not involved in wood processing. These holdings, which
constitute the largest single ownership class, are by and
large concentrated in the North and the South regions where
70 . 8 and 71. 3 , respectively, of timberlands fall in this
category.
TABLE V
DISTRIBUTION OF TIMBERLANDS BY REGION AND OWNERSHIP IN 1977
(MM ACRES)
National Other Forest Other Total Region Forests Public Industry Private % Total
N.W. 16.83 7.52 9.87 7.97 42.19 % 39.9 17.8 23.4 18.9 9.7
Inland: 36.43 6.73 2.09 12.50 57.75 % 63.0 11.6 3.7 21.7 12.2
P.S.W. : 8.17 0.50 2.68 4.94 16.29 % 50.2 3.1 16.4 30.3 3.5
North 9.83 20.69 17.91 117.71 166.14 % 5.9 12.5 10.8 70.8 35.3
South 10.95 6.78 36.24 134.08 188.05 % 5.8 3.6 19.3 71.3 40.0
Total 82.21 42.22 68.79 277.20 470.42 % 17.5 9.0 14.6 58.9 100
Note: totals do not include Alaska and Hawaii. Source: USDA Forest Service (1982), Appendix 3, 344-9.
Distribution of timberlands is an important indicator
of where timber will be produced in the future. From the
standpoint of the near term supply of wood products,
however, supply sources will be determined by the location
45
of existing inventories of mature timber (Dowdle and Henke,
1985 p. 80).
The quantities of timber supply from these sources are
in turn influenced by biological determinants such as land
quality, growth rates, and type of species; and specific
economic objectives pursued by owners in each region.
Attention here will be focused on two general ownership
categories of "public" and "private". These categories
represent the two timber supply sectors for which management
objectves are distinctly defined.
Timber Supply: Private. The economic theory of timber
supply concerns two central issues of when to harvest and
how much to harvest. The problem of when to harvest (optimal
rotation) may be viewed as a special instance in the
classical theory of capital (Hyde, 1980, pp. 60). The profit
maximizing solution to this problem can be derived directly
from the owner's profit function. Under the condition of
perfect competition, and assuming economic rationality,
timber owners choose harvesting schedules that maximize the
present discounted va1~e of their holdings.
In terms of a fixed land-base with an even-aged stock
of timber, the optimality condition implies that trees
should be cut when the discounted marginal product of the
timber stand (annual growth rates) equals the marginal cost
of withholding the timber for an addi tional year. These
costs consist of the interest on the revenues from the cut,
46
value of silvicultural efforts and, as Samuelson (1976) has
pointed out, the rent on the forest land. 3
Economic theories of optimum harvest provide a useful
framework for the analysis of timber supply in the long-run.
Their utility in explaining actual output levels in the
short-run, however, is 1 imi ted. A maj or impediment in the
application of these theories to short-term analysis is the
predominance of noneconomic objectives for private owners
such as tax considerations and the immediate nature of
cash-flow needs. These issues will be considered in greater
detail in the next chapter as we develop empirically
testable relationships for our model.
Timber supply: Public. Conventional economic tools of
price theory (supply and demand analysis) provide an ample
framework for the analysis of timber supply. These economic
considerations will, however, be of limited use in
explaining supply behavior where significant areas of
commercial forestlands are owned, managed, or in one form or
another regulated by public policy. The reason is that in
such cases an entirely different set of administrative
considerations, which are themselves based on biological or
3 For a thourough theoretical treatment of economic issues concerning timber production see Hyde (1980) particularly Chapter 4. A complete mathematical derivation of the optimality condition can be found in Johansson and Lofgren (1985) Chapters 4 and 6.
47
otherwise "non-economic" functions, dominate the amount of
timber made available to the market.
During the course of this century public policy has
become an increasingly important force in the timber market
in the united states. Indeed, as Deacon and Johnson (1985)
have observed, no other natural resource industry has ever
been as heavily influenced by government ownership and
policy as the timber industry. In 1977, the latest year for
which data is available, public ownership accounted for 28.1
percent (135.7 million acres) of the total timberland in the
country (482.5 million acres). Timber production patterns,
at the same time, indicate that in all parts of the country,
particularly in the west, producers of wood products have
become increasingly dependent on timber supplies from public
lands (see Table VI).
Broadly defined, public timber policy encompasses a
wide range of issues concerning management practices,
marketing procedures and harvesting patterns of timber from
publicly owned forests, taxation of private holdings, and
other regulatory policies at the federal and local govern
ment levels. These issues, in particular practices concern
ing the management of the National Forests by the Federal
Government, have been the subj ect of a perennial debate
between economists on the one hand, and environmentalists
and professional forestry experts on the other. A sUbstanti
al body of research has now been accumulated by either camp
TABLE VI
ANNUAL HARVEST OF SOFTWOOD TIMBER BY REGION AND OWNERSHIP (1950-1984)
----------------------------------------------------------------------------------------------------Northwest South Southwest Inland
---------------------- ---------------------- ---------------------- ----------------------Year Total Public Private Total Public Private Total Public Private Total Public Private
(MMCFt. ) " " (HMCFt. ) , " (HMCFt. ) "
, (MMCFt. ) " % 1950 2501.7 22.4 77.6 3424.1 6.1 93.9 731.0 15.0 85.0 537.1 49.8 50.2 1955 2826.0 28.5 71.5 3165.2 7~4 92.6 933.3 21.3 78.7 662.3 71.3 28.7 1960 2744.2 34.8 65.2 2770.4 9.4 90.6 926.6 28.5 71.5 672.6 67.9 32.1 1961 2689.5 37.6 62.4 2697.3 8.8 91.2 915.8 29.9 70.1 687.6 70.3 29.7 1962 2850.6 41.1 58.9 2726.3 9.2 90.8 905.2 30.5 69.5 734.2 72.4 27.6 1963 2957.5 44.8 55.2 2795.8 9.0 91.0 913.2 36.2 63.8 787.0 77.4 22.6 1964 3173.7 45.7 54.3 2981. 3 8.5 91.5 945.0 39.3 60.7 835.1 72.6 27.4 1965 3254.0 45.9 54.1 3121. 3 9.0 91.0 909.8 41.5 58.5 872.8 74.0 26.0 1966 3162.5 42.0 58.0 3240.9 8.3 91.7 903.2 42.9 57.1 896.4 75.5 24.5 1967 3083.4 41. 6 58.4 3243.5 8.4 91.6 855.8 44.2 55.8 893.6 72.2 27.8 1968 3349.7 44.3 55.7 3438.4 8.0 92.0 973.5 48.4 51.6 947.3 75.9 24.1 1969 3099.3 45.0 55.0 3684.7 8.0 92.0 925.8 43.1 56.9 914.6 77.5 22.5 1970 3077.5 37.9 62.1 3690.8 7.6 92.4 904.5 41.2 58.8 831.6 72.0 28.0 1971 3057.6 42.7 57.3 3747.8 7.8 92.2 910.3 45.7 54.3 856.7 71. 5 28.5 1972 3239.3 48.6 51.4 3918.1 8.4 91.6 959.6 46.7 53.3 845.9 69.2 30.8 1973 3275.9 48.5 51.5 3918.8 7.1 92.9 963.5 42.9 57.1 848.4 66.2 33.8 1974 3006.5 41.8 58.2 3829.3 7.9 92.1 778.5 45.3 54.7 749.9 68.0 32.0 1975 2682.1 38.9 61.1 3636.1 9.2 90.8 701.1 44.3 55.7 762.3 60.0 40.0 1976 3113.6 41.7 58.3 3982.4 9.7 90.3 803.7 47.7 52.3 849.6 64.9 35.1 1977 3050.3 40.7 59.3 4218.4 8.7 91.3 820.1 44.9 55.1 873.2 60.5 39.5 1978 3098.5 43.2 56.8 4444.7 9.1 90.9 785.5 46.9 53.1 854.0 58.5 41.5 1979 3046.2 43.6 56.4 4675.2 8.4 91. 6 747.6 47.6 52.4 860.6 58.8 41.2 1980 2673.7 38.8 61.2 4659.2 9.3 90.7 608.1 51.9 48.1 774.7 55.1 44.9 1981 2306.6 37.4 62.6 4770.4 8.7 91.3 526.4 45.2 54.8 715.6 60.1 39.9 1982 2327.3 28.8 71.2 4662.2 7.6 92.4 477.3 44.0 56.0 664.0 50.8 49.2 1983 2727.9 42.2 57.8 2601. 8 17.7 82.3 581. 6 60.4 39.6 681.1 79.7 20.3 1984 2805.3 45.8 54.2 3067.3 15.2 84.8 637.3 51.8 48.2 725.6 71.4 28.6 ----------------------------------------------------------------------------------------------------Source: Appendix A, [1] I [19].
~ OJ
49
in support of its position. 4 Much of this controversy
centers on two principles that have dominated the management
of National Forests in the country over the past two and
half decades: "sustainable yield" and "multiple use", both
embodied in the Multiple-Use sustained-Yield Act of 1960.
A central issue in this controversy concerns the
"sustained yield" requirement of the act and the concepts of
"even flow" and "allowable cut", the practical implications
of the mandate. At the core of criticisms concerning public
timber policy in the U.s. lies the issue of economic
efficiency of these principles and their inconsistency with
the economic concept of "supply".
Regional inventories of softwood sawtimber on National
Forests and other ownerships are presented in Table (VII).
A comparison of figures in Tables (V) and (VII) reveals that
distribution of timber resources does not follow the same
pattern as timberlands. Major differences exist in the
intensity of timber stands across regions and among
ownership categories. The most readily discernible pattern
in the distribution of sawtimber is that the volumes of
standing timber on National Forest lands are generally in
4 Deacon and Johnson (1985) provide a collection of most recent studies on the subject of government timber policy. More technical treatments can be found in Hyde (1980) especially Chapters 2 and 7; and Samuelson (1976).
50
excess of the proportion of forest area in all regions,
indicating much higher densities on the average. This latter
observation has, among other things, been cited as an
important indication that National Forests produce less than
their share of domestic timber output (Howe, 1979 pp. 223).
Region
N.W. %
Inland: %
P.S.W. : %
North : %
South %
Total : %
TABLE VII
VOLUME OF SOFTWOOD SAWTIMBER BY REGION AND OWNERSHIP
1977 (MM BFt.)
National Other Forest Other Total Forests Public Industry Private % Total
386623.0 140322.0 141004.0 59537.0 727486.0 53.1 19.3 19.4 8.2 40.4
260758.2 42774.1 23260.4 53586.8 380379.5 68.5 11.2 6.2 14.1 21.1
157958.0 6356.0 40833.0 50397.0 255594.0 61.8 2.5 16.0 19.7 14.1
8049.9 8.3
11942.1 22734.4 53777.1 96503.5 12.4 23.6 55.7 5.5
33979.7 13883.6 86389.7 206769.5 341022.5 10.0 4.0 25.4 60.6 18.9
847368.8 215277.8 314271.5 424067.4 1800985.5 47.0 12.0 17.4 23.6 100
Note: totals do not include Alaska and Hawaii. Source: USDA For. Service (1982), An Analysis of the Timber situation in the United States, Appendix 3, pp. 370-71.
These and other normative issues cOJ'lcerning public
timber policy are important and merit close scrutiny and
detailed analysis. Such undertaking, however, is well beyond
the scope of the present study. Our interest in the area is
51
a purely empirical one: the manner in which such policies
manifest themselves in timber supply behavior of various
regions; and the insight they can provide in modeling this
behavior.
The extent to which timber offerings from public lands
affect regional markets is determined by the relative share
of the public component of total production. In analyzing
the short-run effects of public timber supply on local
stumpage markets, it is essential to distinguish between
"timber offerings" and actual "harvest levels". Timber from
National Forests and other public lands is usually sold
under four-year removal contracts. Quantities of timber
harvested in anyone period thus do not necessarily
correspond with timber offerings in that period. While
market conditions in the long-run are decided by levels of
timber "sold" from public lands, actual short-run harvest
levels are determined by local demand and price levels. In
short-run analysis, therefore, no distinction need be made
between supply sources. Total harvest, consisting of harvest
on both public and private ownerships, can be explained in
terms of conventional supply theory.
CHAPTER III
THE EMPIRICAL MODEL
In Chapter I we outlined a general conceptual
framework for a mu1tiregiona1, multi-industry model and
described its major ingredients and their conceptual
contents. Based on these theoretical considerations and the
insight we have gathered from our discussions of the lumber
and plywood industries and the timber economy in Chapter II,
we are now in a position to proceed with the development of
our industry-specific model. To aid us in the task, is also
the wealth of information provided by previous econometric
analyses of the forest economy, most notable among them, the
pioneering studies by the Stanford Research Institute
(1954), Holland (1960), Gregory (1960), McKillop (1967,
1969), and other more recent works of Adams and Blackwell
(1973), Robinson (1974), Mills and Manthy (1974), Holley,
Haynes, and Kaiser (1975), Adams (1977), and Adams and
Haynes (1980).
A1 though these studies have, by and large, adopted
fundamentally similar specifications in presentation of
individual behavioral relationships, there are marked
differences among them in terms of objectives, scope (the
degree of spatial aggregation and sectoral presentation),
53
and model structure. 1 From a genealogical point of view, our
model is a direct descendant and extension of Adams'multi-
regional, multiproduct model of the U. S • softwood timber
market (Adams 1977) and the Timber Assessment Market Model
developed by Adams and Haynes (1980).
Both studies were designed as forecasting tools for
the long-range projection of activity in the product and
stumpage markets and assessment of potential impacts of
alternative resource use policies in the private and public
timber sectors. In the taxonomy of regional models, Adams'
approach is a combined top-down/bottom-up framework. The
model considers two products (softwood lumber and plywood),
four product supply regions (including Canadian imports) and
three stumpage supply regions (see Figures 4 and 5 in
Chapter II) • Treating regional supplies as perfect
substitutes, equilibrium quantities and aggregate product
prices (at the national level) are then established jointly
by the levels of aggregate demand and regional supplies. The
model ignores transportation costs so that product prices in
the supply regions are identically the same as the demand
price at the national level.
Timber Assessment Market Model (TAMM) , on the other
hand, is a purely bottom-up model. It is by far the most
1 For a comprehensive survey of econometric studies of the forest economy see Adams and Haynes (1980). A bibliography of earlier projection studies can be found in Gregory, et al. (1971).
54
comprehensive modeling effort in the analysis of the timber
economy in the united states. The spatial structure of TAMM
is based on partitioning of the national market into nine
product and stumpage supply regions and six demand regions.
For each supply region equilibrium solutions for product
prices and quantities are found by simultaneous estimation
of the regional product and stumpage relationships. Given
interregional transportation costs in a given period,
equilibrium spatial price and regional outputs are computed
via a programming procedure which utilizes the reduced form
of the regional supply equations and regional demand
relationships in an iterative process to find the profit
optimizing levels of supply in each region.
The most outstanding feature of TAMM that
distinguishes it from most other projection models of the
forest economy, is its explicitly simultaneous treatment of
interdependent market relationships, specifically, the
linkage between regional stumpage and product sectors which
allows simultaneous determination of market clearing prices
in both markets. For a complete description of the model
structure and interactions of TAMM the reader is referred to
Adams and Haynes (1986).
The model developed in the present study incorporates
and combines features of the two models we have described.
It is, however, designed for the purpose of short term (one
to two year) forecasting. Certain long-run consideraticns
55
such as regional production capacity adjustment nlechanisms
are ignored and resource availability takes on either a
simplified form, or in some cases, is ignored completely.
THE MODEL, ITS STRUCTURE, AND COMPONENTS
The model, as described in Part One, contains three
interrelated components: 1) the national product market, 2)
regional product market, and 3) the regional factor market.
Due to exigencies of estimation, certain modifications
of the initial conceptual model were deemed necessary
without compromising the theoretical consistency of the
model. First, regional lumber supply equations were
estimated in inverse form so that regional lumber prices can
be obtained directly. Second, the regional factor market
relationships were reduced to single equations with price as
the dependent variable. The linkages between the product and
stumpage sectors in each region was thus established by
directly relating prices in the two markets. Third, in the
demand equations for regional lumber outputs, quantities
were replaced by identities that represented actual market
shares. Finally, in the market share equations, a regional
"accessibility index" was included to represent regional
locational advantage. These modifications helped simplify
the model structure considerably. The resulting system, its
components and interactions are illustrated in Figure 9.
The complete system to be estimated may be represented
as:
56
Endogenous Exogenous Aggregate product market: bllQ b12P allxl ( 1)
Regional stumpage market: b24Pp b25Ps a21x2 (2)
Regional product market: b3lQ b32P b33S b34Pp b35Ps a3lx3 (3)
where,
Q and P are, each 2x35 vectors of, respectively, aggregate
quantities consumed and average prices for both lumber and
plywood;
Pp is a 6x35 vector of regional mill level prices of lumber
and plywood;
Ps is a 4x35 vector of regional stumpage prices;
S is a 7x35 vector of supply region lumber and plywood
quantities;
Xl' X2 and X3 are vectors of predetermined variables; and
bll , b12 , b2l , b22 , b3l , b32 , b33 , b34 , b35 , all' a2l , and
a3l are arrays of coefficient estimates for endogenous and
predetermined variables with their appropriate dimensions.
There are four equations in (1), one aggregate demand
and one aggregate supply relationship for each of the two
products. (2) contains four equations, they are regional
stumpage price relationships. There are thirteen equations
in (3); four regional lumber price relationships; four
relationships representing supply of, and demand for, ply
wood in the N. W. and the South; and five regional lumber
market share equations, including Canadian imports.
57
Exogenous ... Aggr. Prod . Variables P" Demand
Import _ ... Product Prices r- Imports ....
~,
Aggr. Prod. ...... ... Aggr. Prod • Supply """ ~ Price
~
product~vi~ Reg~onal ......
Rates ~ Prod. Supply
Wage ..
I Inventory ~---. , Regional .... _ .. Regional
stump. Price .... r- Prod. Price Inst~tution-~ al Factors
~ ..-Regional
~ Relative
Exports Reg. Prices
Transport .. Regional Costs r Mkt. Shares .. ..
Figure 9. Model structure and its main interactions.
Following a general explanation of the estimation
procedures and presentation of results, we will describe
each of these model components separately and in detail;
and will provide theoretical justifications for the forms
used in the final estimating structures.
58
MODEL ESTIMATION
As a first approximation, all parameters of the
behavioral equations in the model were estimated via
Ordinary Least Square (OLS). This was done to ascertain the
appropriateness of specifications; and to test the goodness
of fit in individual relationships. In several cases this
procedure led to some modification in the equations under
study.
The model contains several interdependent relation
ships. Two Stage Least Square procedure was therefore used
as the estimation method for all relationships where one or
more of the explanatory variables were endogenously deter
mined. Since the model contains a large number of exogenous
variables, in order to avoid problems with degrees of
freedom, it was necessary to select only a subset of
predetermined variables as instruments in the first stage.
Traditionally, the problem of choice among predeter
mined variables has been dealt with either through arbitrary
selection of a small set of regressors, or by using a
principal components method to derive linear combinations of
predetermined variables that maximizes their covariance (for
the original presentation of the latter approach see Kloek
and Mennes, 1960). The first procedure is obviously faulty
and can lead to serious errors in specification. The second
procedure, though it has been shown to produce efficient
59
estimates in large samples, can cause confusion in the
theoretical interpretation of the mixture of variables.
In order to maintain procedural consistency in
selection of first stage regressors, a method was adopted
based on guidelines suggested by Fisher (1965) and McCarthy
(1971). The procedure was as follows: for each explanatory
endogenous variable, a set of regres30rs was selected that
include: 1) all predetermined variables in the equation
containing the endogenous variable, 2) all exogenous
variables in the equation that explains that variable, and
3) any lagged values of the endogenous variable present in
the model. Additional first stage regressors were then drawn
from other parts of the model by causal ordering of
variables in the entire system and choosing only variables
of the first causal order, i.e., variables that were judged
to be directly related to the equation under study.
An additional complication in the estimation of the
model arises from the contemporaneous correlation between
the disturbance terms in the lumber market share equations.
This problem emerges from the implied restriction that the
market shares must sum to unity, or 100 percent, so that in
any given period an over (under) estimation of one region's
market share is bound to lead to under (over) estimation of
other regions' market shares. To deal with this problem, the
lumber market share equations were estimated as a separate
block by the more appropriate method of Three stage Least
60
Square (Zellner, 1962; Zellner and Theil, 1962).2 The
decomposition of the model in this fashion is justified
because no current endogenous variables appear on the
right-hand side of these equations. Verification by the
order condition shows that the identification requirement of
the Zellner procedure is also satisfied. In the NW market
share equation this procedure resulted in an unreasonably
small coefficient on the accessibility index. These
equations were subsequently reestimated with imposed
restriction that the coefficient on the accessibility index
(see Table XIV) for NW should equal the value obtained from
the OLS estimate. The estimation results of the complete
model are presented in Table VIII. For the definitions of
variables, their derivation and sources see Table IX.
TABLE VIII
PARAMETER ESTIMATES OF THE MODEL
(1) Lumber Demand (Aggregate Consumption) TLUMCON = 34867.25 - 64.07*LUMPRUS - 31050*MAL/PLY +
(2.5) (4.1) 9.19*PISSPRD + 10.28*MAHSTRT + 21.20*PERSINC (1.6) (6.5) (2.7)
DW = 1.64 Estimation Method: 2SLS 2 R = .842
(2) Lumber Consumption Identity TLUMCON = TLUMPRD + TLUMIMP - TLUMEXP
2 The method is based on simUltaneous solution of all parameters of the system using the variance-covariance of the disturbance terms of equations. For a complete description of the method see Johnston (1984, pp. 486-90) and Pindyck and Rubinfeld (1981, pp. 334-6).
61
(3) Average Price of Lumber in u.s. LUMPRUS = -37.16 + 0.238*LUMPRUS_1 + 0.000994*TLUMCON +
(1.93) (1.72) 1.665*STMPRUS + 19.77*RW/ILPL (3.3) (4.0)
R2 = .844 DW = 1.30 Estimation Method: 2SLS
(4) Plywood Demand (Aggregate consumption) TPLYCON = - 7356.4 - 8.30*PISWPLY - 183.1*MAPLY/S +
(1.1) (0.2) 154.68*VNCONST + 72.57*INDINDP (3.5) (1.8)
R2 = .967 DW = 1.56 Estimation Method: I2SLS
(5) Plywood Supply TPLYPRD = 7105.7 + 4.14*PISWPLY - 2656.0*RW/ILPP +
(1.05) (2.4) 0.699*TPLYCON
R2 = .955 DW = 1.51 -lstimation Method: 2SLS
(6) Stumpage price NW STMPRNW = 47.20 + 0.005*TTCUTNW_1 + 0.413*LUMPRNW -
(1.23) (6.0) 0.0015*INVPRNW - 0.0047*TVSNFNW (4.7) (0.61)
R2 = .861 DW = 1.38 Estimation Method: 2SLS
(7) stumpage Price so STMPRSO = 0.954 + 0.988*STMPRSO_1 + 0.654*LPCHGSO
(11.1) (6.5) R2 = .803 DW = 1.88 Estimation Method: 2SLS
(8) stumpage Price SW STftPRSW = -11.72 + 0.0092*TTCUTSW_1 + 0.485*LUMPRSW -
(2.0) (13.8) 0.00087*INVPRSW - 0.021*TVSNFSW (2.5) (2.3)
R2 = .955 DW = 2.2 Estimation Method: 2SLS
(9) stumpage Price IL STMPRIL = 1.86 + 0.853*STMPRIL_1 + 0.235*LPCHGIL
(13.3) (7.1) R2 = .873 DW = 2.31 Estimation Method: 2SLS
(10) Price of Lumber NW Region LOMPRNW = -33.9 + 0.26*LUMPRNW_1 + 0.0056*LUMPDNW +
(2.0) (4.9) 0.917*STMPRNW + 4.09*LCSTLNW
(5.6) (1.7) R2 = .829 DW = 1.63 Estimation Method: 2SLS
62
(11) Price of Lumber SO Region LUMPRSO = -19.9 + 0.339*LUMPRSO_l + .0046*LUMPOSO +
(2.9) (3.5) 0.73*STMPRSO + 0.72*IRLWRSO - 0.26*ILPOLSO
(0.6) (4.4) (3.8) R2 = .760 OW = 1.65 Estimation Method: 2SLS
(12) Price of Lumber SW Region LUMPRSW = 26.51 + O.23*LUMPRSW_
l + 0.0039*LUMPOSW +
(3.2) (1.65) 1.34*STMPRSW + 0.03*IRLWRSW
(7.6) (1.23) R2 = .960 OW = 1.98 Estimation Method: 2SLS
(13) Price of Lumber IL Region LUMPRIL = 1.55 + 0.286*LUMPRIL_
l + 0.0029*LUMPOIL +
(2.4) (.62) 1.29*STMPRIL + 12.8*LCSTLIL
(2.4) (2.7) R2 = .805 OW = 1.83 Estimation Method: 2SLS
(14) Plywood Supply NW PLYPDNW = 2173.1 + 0.962*PLYPONW_ l + 13.3*PLYPIOF -
(10.1) (1.0) 33.13*STMPRNW - 834.7*LCSTPNW (1.5) (1.3)
R2 = .904 DW = 1.74 Estimation Method: 2SLS
(15) Plywood Supply SO PLYPDSO = -1999.7 + 0.742*PLYPOSO_ l + 18.05*PLYPISP +
(9.3) (2.5) 4.87*ILPOPSO
(1. 42) R2 = .981 DW = 2.16 Sample 1965-84, OLS
(16) Market Share Lumber Canada MKTSLCA = 9.237 + 0.938*MKTSLCA_l - 8.886*MACA/US
(14.4) (1.4) R2 = .967 OW = 1.30 Estimation Method: OLS
(17) MKTSLCA = 100 * (LUMIMPC / TLUMCON)
(18) Market Share Lumber NW MKTSLNW = 1.27 + 0.934*MKTSLNW_l + 0.218*(LPUS-NW) +
(12.4) (2.3) 0.095*ACCSLNW
2 (--) R = .914 OW = 2.1 Estimation Method: 3SLS
(19) MKTSLNW = 100 * [(LUMPONW - EXPLUNW) / TLUMCON]
(20) Market Share Lumber SO MKTSLSO = 9.29 + 0.0017*LUMPDSO_1 + 0.152*(LPUS-SO) +
(7.0) (1.9) 0.0082*ACCSLSO
2 (1.4) R = .610 DW = 1.36 Estimation Method: 3SLS
(21) MKTSLSO = 100 * [(LUMPDSO - EXPLUSO) / TLUMCON]
(22) Market Share Lumber SW
63
MKTSLSW = 7.82 + 0.00147*LUMPDSW_1 + 0.149* (LPUS-SW) + (7.1) (9.0)
0.00876*ACCSLSW 2 (2.0)
R = .923 DW = 1.58 Estimation Method: 3SLS
(23) MKTSLSW = 100 * [(LUMPDSW - EXPLUSW) / TLUMCON]
(24) Market Share Lumber IL MKTSLIL = 1.69 + 0.815*MKTSLIL_1 + O.0667*(LPUS-IL) +
(9.6) (1.5) O.00065*ACCSLIL
2 (0.9) R = .738 DW = 2.42 Estimation Method: 3SLS
(25) MKTSLIL = 100 * [(LUMPDIL - EXPLUIL) / TLUMCON]
(26) Demand for NW Plywood PLYPDNW = 15369.1 + O.239*TPLYCON - 9658.1*MANW/SO +
(5.5) (11.9) 149. 62*ACCSPNW
2 (3.4) R = .926 DW = 1.97 Sample 1965-84, 2SLS
(27) Demand for SO Plywood PLYPDSO = 5784.9 + 0.719*TPLYCON - 16289.8*MASO/NW +
(8.0) (7.4) 184.4*ACCSPSO
2 (1.4) R = .936 DW = 1.51 sample 1965-84, 2SLS
Figures in parantheses are t statistics.
TABLE IX
DESCRIPTION OF VARIABLES USED IN THE MODEL
ACCSLNW Accessibility index for NW lumber, defined in text.
ACCSLSO Accessibility index for SO lumber.
ACCSLSW Accessibility index for SW lumber.
ACCSLIL Accessibility index for IL lumber.
ACCSPNW Accessibility index for NW plywood.
ACCSPSO Accessibility index for SO plywood.
EXPLUNW Exports of lumber from NW (MMBF) [2, 3, 18].
EXPLUSO Exports of Southern Pine (MMBF) [3, 4, 18]
EXPLUSW Exports of lumber from California ports (MMBF) [2, 3].
EXPLUIL Exports of lumber from IL (MMBF) [2, 3, 4,].
64
I LPDLSO Index of labor productivity (MBF per man-year) in lumber production in SO (1967=100) [5, 6].
ILPDPSO Index of labor productivity (MFt. 2 per man-year) in plywood production in SO (1967=100) [5, 7].
INDINDP Index of industrial production (1967=100) [8].
IRLWRSO Index of hourly earnings in lumber and wood products industries SO (1967=100). See text for derivation [9].
IRLWRSW Index of hourly earnings in lumber and wood products industries SW (1967=100). See text for derivation [9].
INVPRNW Total private inventory of timber (forest industry and other privale holders) at the start of the period NW. (MMFt. ) [1].
INVPRSW Total private invento3y of timber at the start of the period SW. (MMFt. ) [1].
LCSTLNW Labor costs in lumber production NW. Derived by dividing real wage rates by the index of labor productivity in NW sawmills [9, 5, 6].
LCSTLIL Labor costs in lumber production NW. Derived by dividing real wage rates by the index of labor productivity in IL sawmills [9, 5, 6].
65
LCSTPNW Labor costs in lumber production NW. Derived by dividing real wage rates by the index of labor productivity in NW plywood mills [9, 5, 6].
LPUS-NW (LUMPRUS-LUMPRNW).
LPUS-SO (LUMPRUS-LUMPRSO).
LPUS-SW (LUMPRUS-LUMPRSW).
LPUS-IL (LUMPRUS-LUMPRIL).
LPCHGSO (LUMPRSOt - LUMPRSOt _1)·
LPCHGIL (LUMPRILt - LUMPRILt _1) •
LUMIMPC Lumber imports from Canada (MMBF) [1] •
LUMPDNW Lumber production in the NW (MMBF) [1] •
LUMPDSO Lumber production in the SO (MMBF) [1] •
LUMPDSW Lumber production in the SW (MMBF) [1] •
LUMPDIL . Lumber production in the IL (MMBF) [1] . . LUMPRNW Price of softwood lumber NW, f.o.b. mill, $/MBF
(Deflated by All Commodity PPI) [1].
Ltn{PRSO Price of softwood lumber SO, f.o.b. mill, $/MBF (Deflated by All Commodity PPI) [1].
LUMPRSW Price of softwood lumber SW, f.o.b. mill, $/MBF (Deflated by All Commodity PPI) [1].
LUMPRIL Price of softwood lumber IL, f.o.b. mill, $/MBF (Deflated by All Commodity PPI) [1].
LUMPRUS Weighted average f.o.b. mill price of all producing regions including Canadian imports, deflated by All Comma PPI.
MACA/US Two-year moving average ratio of price of Canadian lumber to average price of lumber in U.S., both prices in U.S. currency [1, 10].
MAHSTRT Two-year moving average of housing starts in U.S. [11, 12].
MAL/PLY Two-year moving average ratio of Producer Price Index for softwood lumber to the Producer Price Index for softwood plywood.
66
MAPLY/S Two-year moving average ratio of PPI for softwood plywood to composite index of sUbstitutes: PISWPLY /(.S*PPI particle boards + .2S*PPI gypsum products + .2S*PPI building paper and board) [13, 14].
MANW/SO Two-year moving average of (PLYPIDF / PLYPISP).
MASO/NW Two-year moving average of (PLYPISP / PLYPIDF).
MKTSLCA Canadian lumber imports as percent of total u.s. consumption. [1]
PERSINC Aggregate national personal income (MMM$), deflated by GNP Implicit Price Deflator (1972=100) [15].
PISWPLY Producer price index for all softwood plywood (1967=100) [13, 14].
PISSPRD Producer price index for structural steel products (1967=100) [14].
PLYPDNW Plywood production in the NW (MMBFt.2) [1].
PLYPDSO Plywood production in the so (MMBFt.2) [1].
PLYPIDF Price of Douglas fir plywood, deflated by All Comm. PPI. (Index 1967=100) [1].
PLYPISP Price of Southern pine plywood, deflated by All Comm. PPI. (Index 1967=100) [1].
RW/ILPL Real hourly earnings in lumber and wood products industry in u.s. divided by the index of output per employee in sawmills [9, 16].
RW/ILPP Real hourly earnings in lumber and wood products industry in u.s. divided by the index of output per employee in plywood mills [9, 16].
STMPRNW Average price of timber harvested from National Forests in the NW, $/MBF, (deflated) [1].
STMPRSO Average price of timber harvested from National Forests in the SO, $/MBF, (deflated) [1].
STMPRSW Average price of timber harvested from National Forests in the SW, $/MBF, (deflated) [1].
STMPRIL Average price of timber harvested from National Forests in the IL; $/MBF; (deflated) [1].
STMPRUS Weighted average price of timber harvested from National Forests in all regions, 1967 $/MMBF.
TLUMCON Total apparent domestic consumption of softwood lumber, MMBF [1].
67
TLUMEXP Total annual exports of softwood lumber, MMBF [1].
TLUMIMP Domestic imports of softwood lumber from all origins, MMBF [1].
TLUMPRD Total domestic lumber production, MMBF [1].
TPLYCON Total apparent2domestic consumption of softwood plywood, MMFt. •
TTCUTNW Total annual3harvest of timber by all ownerships in NW, MMFt. •
TTCUTSW Total annual3harvest of timber by all ownerships in SW, MMFt. [ 1] •
TVSNFNW Total annual volume of timber3auctioned from National Forests in NW, MMFt. [I, 19].
TVSNFSW Total annual v9lume of timber3auctioned ~rom National Forests in SW, MMFt. [I, 19].
VNCONST Total value of new construction put in place, deflated by the GNP Implicit Price Deflator (1972=100) [17].
* Numbers in brackets indicate sources of data as listed in Appendix A.
3 All volumes of timber auctioned and sold from National Forests were converted from BF, local log rules to cubic feet. The appropriate conversion factors were provided by Dr. R.W. Haynes at the PNW Forest and Range Experiment station in Portland, Oregon. These conversion factors are: Western Oregon and Washington: 5.5~ Eastern Oregon 2nd Washington: 4.96; SO: 4.67; SW: 5.27; IL: 4.67 BF per Ft ••
68
MODEL COMPONENTS
The Aggregate Product Market
The four equations in this part of the model represent
a simple supply-demand relationship for each of the two
products. The postulated relationships in these equations
are essentially the same as the conceived structures set
forth in Chapter I. Their specifications are common to
several other econometric studies of the wood products
industry, as well as other commodity models.
In the demand equations, quantities of aggregate
product demand (apparent consumption) are related to own
price, price of sUbstitutes and one or more measures of the
level of activity in relevant end-use markets. The choice of
the latter variables, sometimes referred to as "shifters"
(since they serve to locate the function), is based on our
discussion of end use markets in Chapter II.
Another component of the demand relations is the
substitution effect. The problem of correctly identifying
patterns of SUbstitution among commodities is a serious one
and merits careful consideration. Forces of technological
change are, more often than not, quite unpredictable. New
products appear on the market on short notices; and old ones
disappear from the face of the market unnoticed. Indeed, as
Pringle (1971) has pointed out, many cases of substitution
in the forest products market have been the result of
development of new and often superior products, especially
within the industry itself. In such cases, reliance on the
69
conventional price considerations to explain substitution,
may be inappropriate. Due to the time related nature of
these developments, the time dimension in demand analysis,
therefore, becomes crucial. The longer the period of
analysis, the higher the likelihood of error.
In our analysis, plywood and structural steel prices
are included in the lumber demand relation to account for
the SUbstitution effect. In case of plywood, the moving
average ratio of plywood price index to a composite index of
three substitutes is used (see Table IX). In both cases the
commodities considered are established products with a long
history of competition with the product under study.
The aggregate product supply functions are also
conventional in form, except that the lumber supply equation
is expressed in terms of price. The average price of lumber
in the national market is determined by its own value in the
previous period, total quantity consumed, average price of
stumpage from all regions, and real wage rates (hourly
earnings) in the industry. In order to account for the
offsetting influence of historical gains in productivity on
the rising cost of labor, wage rates in both relations are
adjusted by the BLS indexes of labor productivity for saw
mills and planing mills (SIC 2421) and plywood (SIC 2430) •
Regional stumpage Markets
Direct estimation of structural relationships in the
regional stumpage markets posed many serious problems.
70
Differential patterns of ownership among regions, diversity
of economic objectives pursued by the private holders within
each region, and combination of various species in each
geographical area are but a few of the factors contributing
to the complexity of this market. It is therefore not
surprising that the majority of previous empirical studies
of the timber economy have resorted to somewhat ad hoc
specifications in explaining the behavior of this market.
The approach adopted here begins with the stipulation
that the derived demand behavior for stumpage is more
appropriately described in a dynamic setting. Firms
producing the final products, i.e. lumber and plywood,
usually do not plan their levels of output in the spur of
the moment. Adjustments in capacity and input flows, and
employment decisions, produce rigidities in their supply
adjustments that require some advance planning (Adams 1977,
p. 20). In addition, due to the relatively low short-run
price elasticities of demand for lumber and plywood, much of
the increase in factor price (i.e. stumpage) can be
transmitted to the final product market. It is therefore
unlikely that producers' demand for stumpage would adjust
immediately to changed conditions of price.
We can therefore postulate that the "expected" level
of demand for stumpage is given by:
(3.1)
71
where Pt and Plt are, respectively, current stumpage and
lumber prices. However, due to the constraints discussed
above, we would expect the actual demand to adjust only
partially to the expected level, that is:
(3.2)
which, by substitution and rearrangement of coefficients to
estimable forms, gives the following estimating relationship
that is indeed the partial adjustment model widely appli9d
to agricultural commodities (Nerlove, 1958a, 1958b):
(3.3)
On the supply side, we assume that harvest levels are
determined by stumpage prices and other supply determinants.
In our analysis we relate stumpage supply directly to price,
and inventory levels at the start of the period (INV). Also,
we include timber offerings from the National Forests (QNF)
as an additional shift variable so that the stumpage supply
relationship is expressed as:
(3.4)
Equations (3.3) and (3.4), together with the implied
equilibrium condition, provide all the information we need
on the regional stumpage markets. Given these simultaneous
72
relationships for each region, it is possible to obtain an
estimating relationships for stumpage price directly from
the reduced form of equations (3.3) and (3.4) by regressing
Pt on all the exogenous variables in the system.
In estimation, the above formulation produced
satisfactory results only for the NW and SW markets. In the
case of the Southern stumpage, the coefficient for the
National Forest offerings, contrary to our expectation,
carried a positive sign. In the Inland region, on the other
hand, although the coefficients all had the right signs, two
were found to be statistically insignificant. An even more
serious problem in the case of the Inland region proved to
be the presence of very high serial correlation and the fact
that corrective procedures led to incorrect signs on two
coefficients. These results are reported below. Note that
the inventory volume for the Inland region has been replaced
by the ratio of total private harvest to private inventory.
(3.5) STMPRSO = -37.09 + 0.482*LUMPRSO + 0.0685*TVSNFSO + (5.9) (7.5) (1. 8)
0.0064*TTCUTSO_1 - 0.0957*INVPRSO (3.6) (0.7)
2 _ R - 0.871 OW = 1.69
(3.6) STMPRIL = -6.39 + 0.296*LUMPRIL - 0.0008*TVSNFIL + (29) (6.6) (0.19) 0.02l*TTCUTIL_1 + 0.224*CUT/IIL (2.7) (0.08)
R2 = 0.777 OW = 0.944
73
Due to these difficulties an alternative approach is
adopted. It is based on a simplified derived demand approach
and is a variant of the price mark-up procedure applied to
the analysis of many agricultural commodity prices (Tomek
and Robinson 1952) and the stumpage price projections in the
USDA 1982 analysis of the timber situation in the United
states (USDA/FS 1982, pp. 150 and 212).
In its most simplified and general form, the method
invo1 ves the establishment of a relationship between the
product market and the factor market in terms of a marketing
margin:
(3.7) PP = FP + MM
or inversely,
(3.8) FP = PP - MM
where PP is the product price, FP is the factor price, and
MM denotes the marketing margin.
Haynes (1977) has suggested several possible forms for
the linkage between the two markets based on alternative
treatments of the marketing margin as either constant, a
constant percentage, or a linear function of prices. In the
case of the latter, stumpage prices are regressed on product
prices and the coefficient on the product price is
74
interpreted as the coefficient or elasticity of price
transmission between the two markets according to whether
the prices are expressed in logarithms or not. In our study
the relationship between stumpage price and lumber price
follows a specification that has come to be known as the
"pass-through" technique in the forestry literature (Haynes
ibid., p.284). It involves an expression relating the price
change in the two markets. The implication here is that
changes in the product market are only partially transmitted
to the stumpage market; i.e.
(3.9)
This relationship was used to estimate stumpage prices in
all four regions. As indicated by the statistical properties
of the estimates shown in Table X, the relationship appears
to adequately depict the linkages between regional stumpage
and product prices in all regions. Note that the
coefficients on lumber price change measure the extent to
which changes in product prices are transmitted to the
stumpage market, and thus reflect the elasticity of supply
in regional stumpage markets. The results in Table X
indicate that for a shift in demand for regional lumber,
greater stumpage price changes in the SO than in the NW
would be expected. The elasticity of regional stumpage
supplies has been examined in several other studies such as
Adams (1976), Haynes (1977), and Adams and Haynes (1980). A
75
comparison of findings from these studies indicate that the
stumpage market in the SO has lost much of its fluidity over
the past two decades. In the light of relatively significant
declines in NW stumpage prices during the past ten years, it
is not surprising to find a higher elasticity in the NW
stumpage market than in the so.
The results of this latter formulation were used to
represent the product-stumpage linkage in the South and
Inland regions. No attempts were made to incorporate the
inventory and National Forest offerings in this formulation.
TABLE X
ESTIMATION RESULTS FOR REGIONAL STUMPAGE PRICE RELATIONSHIPS
STMPRNW = 2.55 + 0.932*STMPRNW_1 +
2 (15.6) R = 0.889 OW = 1.96
STMPRSO = 0.95 + 0.988*STMPRSO_ l +
2 (ll.l) R = 0.803 OW = 1.88
STMPRSW = 2.30 + 0.909*STMPRSW_1 +
2 (17.3) R = 0.920 OW = 2.3
STMPRIL = 1.86 + 0.853*STMPRIL_ l +
2 (13.3) R = 0.873 OW = 2.04
0.281(LUMPRNW - LUMPRNW_l
) (4.6)
0.654(LUMPRSO - LUMPRSO_l
) (6.5)
0.454(LUMPRSW - LUMPRSW_l
) (9.8)
0.235(LUMPRIL - LUMPRIL_l
) (7.1)
The reason is that although both variables are important
indicators of supply behavior in the long-run, they are
unlikely to resu1 t in the loss of much information in a
76
short-run analysis. Their omission, however, does somewhat
limit the utility of the model in the area of policy
analysis. This is, nevertheless, a small loss within the
context of the model as a whole.
A complete analysis of the regional stumpage markets
requires a different and more appropriate setting than the
present one. The impact of the Forest Service policies
regarding harvest levels, for instance, can more amply be
analyzed within the framework of regional private stumpage
markets. For, as Haynes and others (1981), and Connaughton
and Haynes (1983) have shown, the relative impact of changes
in the National Forest offerings on local stumpage prices
depend largely on the behavior of private owners; and their
reaction --or ability to react-- to the changing price
conditions. In other words, if the stumpage supply from
private ownerships is elastic, then we can expect the effect
of changes in National Forest offerings to be mitigated by
adjustments in the private supply.
In order to examine this proposition, price
elasticities of private stumpage supply were estimated for
the four regions. The analysis is based on a simple
regression equation that relates supply of stumpage from all
private ownerships ("forest industry" and "other private"
owners) in a given region to the price of stumpage and
inventory levels in that region. The results are tabulated
in Table XI. The aim of the analysis is purely illustrative.
It is intended to measure possible losses of information
77
incurred as a result of omissions in our approach to
modeling the stumpage price in the SO and IL regions. Though
the statistical properties of these estimates were in
general satisfactory, the resulting low coefficients of
determination (between .57 to .79) and serial correlation in
some cases, precluded the use of these relationships in our
model.
The observed variations in cross regional supply
behavior in the private stumpage markets, as indicated by
these elasticity measures, is considerable. stumpage price
elasticities in the NW and so are shown to be higher than
those found in other regions. Inventory, on the other hand,
appears to be a much more important factor in determining
supply patterns in the SW and IL regions. These results are
generally consistent wi th those found by previous studies
for example Adams (1977), Adams and Haynes (1980), and
Adams, et.al. (1982). One important implication of these
TABLE XI
ESTIMATES OF REGIONAL ELASTICITIES OF PRIVATE STUMPAGE SUPPLY
. . . * Elast~c~t~es
supply Region Stump. Price Inventory
NW 0.493 0.732 SO 0.458 0.162 SW 0.135 1.564 IL 0.111 1.073
* Point estimates measured at means for each variable.
78
results is that the impact of public timber offerings will
probably be much smaller in the SO and NW than in the sw and
IL regions. with regards to the inventory levels, on the
other hand, it seems that resource availability is a much
more constraining factor for the suppliers in SW and IL than
in NW, and especially the SO region. Empirical evidence of
rapid expansion of private inventories in the South during
the sample period also support this conclusion.
Regional Product Markets
This part of the model contains a total of thirteen
equations: four relationships for regional lumber prices,
four relationships for plywood supply and demand in NW and
SO, and five regional market share relationships. The
regional lumber price and plywood supply relations are
specified in forms identical to the aggregate market
relations. All equations include labor productivity,
expressed in terms of physical output per man-year in each
period. It is calculated by dividing annual regional output
by the average number of production workers. Inadequate
regional or state level data precluded the construction of
more suitable measures such as value added per employee. The
productivity measure is employed here as a surrogate for
technological change in the industry over time. See Appendix
C for data.
Technological change is defined here in the positive
sense of the development or adoption of more efficient
79
production techniques such that greater output can be
obtained from a given combination of inputs. But this
conclusion (technological change) cannot be inferred
directly from the data. An alternative --and equally valid--
explanation of the movements in this measure is economies of
scale. In other words, the over time variations in the
output/employment ratio may be interpreted as actual shifts
in the production function, or movements along the same
function. If the latter is indeed the case, then we can
expect the productivity measure to be highly correlated with
output and lead to collinearity problems in estimation of
supply functions.
We have therefore used this measure with caution. In
all cases where the index of productivity exhibited a high
correlation with output, it was either excluded from the
relationship, or used as a deflating factor to adjust real
wage rates.
Regional wage rates in the lumber and wood products
industry (SIC 24) were calculated as the average of hourly
earnings in the industry in the major producing states in
that region. 4 This data, however was not available for the
entire sample period. Industry-specific data on hourly
4 The regional averages are based on data from a select number of leading states. They are; NW: Oregon and Washington; so: Texas, S.Carolina, Florida, Georgia, Luisiana; SW: California; IL: Idaho, Colorado, Montana, Utah, and N.Mexico.
80
earnings at the state level are generally available only
from 1972 as reported by the Bureau of Labor statistics. The
remainder of the series from 1950 were estimated by first
constructing an index of average regional hourly earnings in
manufacturing (with 1967 as the base) from the state level
data; and then applying this index to the first available
observation in the series to calculate the rest of the
observations. These data are reported in Appendixe B.
In order to check the accuracy of these estimates, a
second procedure was adopted for deriving the same
estimates. This latter procedure consisted of adjusting
weighted regional average hourly earnings in manufacturing
by the weighted ratios of hourly earnings in manufacturing
to average hourly earnings in the lumber and wood products
industry at the national level in each period; i.e., for
each region,
(3.9) EMPLWPt*AVWRLWPt EMPLWPUSt*AVWRLWPUSt = EMPMFGt*AVWRMFGt EMPMFGUSt*AVWRMFGUSt
where the first two terms are respectively regional wage
rates in lumber and wood products, and manufacturing; and
the terms on the right-hand side are the same variables at
the national level. The results from both procedures were
quite close and produced identical results in estimation.
The two regional plywood supply equations were
specified according to the classical partial adjustment
81
model and include a one-period lagged quantity term, own
price, and factor prices as explanatory variables.
Estimation of the plywood equation for the Southern
plywood produced a positive sign on the stumpage price vari
able. The stumpage price term was consequently dropped in
final estimation. This result, though contradictory to the
postulated relationship, may have an explanation. Softwood
plywood is a relatively new product in the Southern market
and has expanded very rapidly during the past decade. Expan
sion of plywood output has in turn intensified local com
petition for stumpage, leading to higher prices in the
stumpage market. Hence, the positive correlation is
justified; yet the stumpage price term cannot be retained
because the causal inference with respect to the direction
of causation, as implied by the supply equation, is no
longer valid.
Regional Market Shares
The seven equations in this part of the model
constitute the pivotal relationships in our model and
comprise the very core of the analysis. structurally, all
other components of the model can be viewed as peripheral to
this set in the sense that the information they provided is
ultimately channeled into this set of relationships.
These equations are all essentially similar in form
and structure; except that the two plywood share equations
are expressed in terms of total regional output. Regional
82
market shares in plywood are therefore obtained indirectly
from the ratio of output estimates to total consumption at
the national level. In estimation, regional exports of
plywood were ignored because they constitute a small
percentage of regional production and the volumes are very
close in both regions. The plywood market share equations
were estimated for both regions over the sample period
1965-1984. The reason is that the advent of plywood
production in the South represents a major transformation in
the competitive structure of the plywood market. Since
practically no plywood was produced in the South before
1964, the behavior and position of the NW prior to this date
would be of little relevance to the analysis of competition
in this market.
Each equation contains a measure of price competitive
ness which is expressed either in terms of price ratios (in
plywood equations), or price difference (in lumber
equations). In the market share equation for Canadian
imports, the price of Canadian lumber is measured in u.s.
dollars. Thus the impact of fluctuations in currency
exchange rates are incorporated in the price term. All
market share relationships for lumber contain a lagged
output term. The underlying justification for this
formulation is that changes in interregional trade patterns
do not respond instantaneously to small changes in market
conditions. In view of the relatively low elasticities in
regional supply and demand conditions that prevail in the
83
lumber and plywood markets, this is indeed a very reasonable
assumption. In the lumber market share equations for the NW
and IL regions the lagged output terms were replaced by the
regional market shares in the previus period. Since in both
regions, the NW in particular, exports constitute a
sUbstantial share of the total output, the use of the lagged
output term appeared inappropriate in that they lead to over
estimation of actual shares of the domestic market.
The other major component of the market share
equations is the regional accessibility index which
represents relative locational advantage of each region.
These measures are intended to approximate the overall
economic accessibility, or, the potential of each region in
marketing its outputs. Construction of this index is based
on the principle of "potential interaction" as employed in
the context of general gravity models.
Gravity models constitute a large family of flexible
methods widely applied to a range of analytic situations
involving interaction between spatially separated nodes. The
general approach in gravity models is predicated on the
physical law of mass gravitation. It explains gravitation
(or interaction) between two points in space in terms of
some surrogate measure of their respective masses, and as a
decreasing function of the distance between them. Among the
many different forms of the gravity model used in trans
portation and migration analysis, the types that are of
84
relevance to the present study are the commodity flow 5 variants. As distinct from the more general forms of the
gravity models that measure interaction "between" two
points, commodity flow models are based on the "potential"
concept of gravity that pertains to the measurement of
potential interaction at a given area or point. Within the
context of commodity flow analysis, the most general
formulation of the potential gravity model may be given as:
(3.10) F .. = bO. f(d .. ) 1) ) 1)
where Fij is the commodity flow from i to j, OJ is a measure
of demand at j for the commodity under question, d.. is 1)
distance or a measure of transport cost between the two
points, and b is a constant. 6
Our presentation of regional accessibility is based on
a formulation of the gravity model given by Isard (1960 p.
499). It is defined as:
5 Attempts at formalization and theoretical derivation of the gravity model have followed two distinct paths: the probabilistic approach of Isard (1960, pp. 493-502) and Wilson (1967); and the economic approach based on the utility theory set forth by Niedercorn and Bechdolt (1969). For a comprehensive treatment of the gravity model see Isard, ibid. pp. 493-566.
6 This formulation and parts of the discussion are borrowed from Richardson (1979 pp. 194-5).
85
(3.11)
where Ai is accessibility of the supply region i, Dj denotes
level of demand in region j, and TC.. is the average ~J
transport cost between i and j for all demand regions. Two
different indexes of accessibility were constructed using
ei ther personal income or annual change in population to
approximate levels of regional demand for lumber and ply
wood. Both variables are highly correlated with the level of
construction activity and are potentially appropriate
surrogates for wood consumption. The second index, measured
in terms of population change, however, produced
consistently better statistical results in all equations and
was adopted. In regional market share equations for lumber,
the exponent to be applied to transportation costs was set
to equal one. An alternative formulation of (3.11) above,
wi th an exponent of two on transport costs ( a=2) produced
much better results in the plywood demand equations and was
consequently used. The obtained index numbers are shown in
Table XIV.
Interregional transport costs were derived for every
period in the sample by applying the Producer Price Index
for Wood Products Freight to the average 1977 estimates of
transport costs for lumber and plywood provided by the
Pacific Northwest Forest and Range Experiment station. See
Tables XII and XIII.
86
since the Bureau of Labor statistics began reporting
the Producer Price Index for Wood Products Freight in 1969
(see Fehd 1975), it was necessary to construct the remaining
series from 1950 indirectly. This was accomplished by
regressing the available observations on the Producer Price
Index for Transportation Services during the same sample
period. The missing observations were thus estimated from
the regression equation: PPIWPF = -32.571 + 1.265 [PPI
Trans. services] (R2=.99). See Appendix D.
From/To
NW
SO
SW
IL
From/To
NW
SO
SW
IL
TABLE XII
INTERREGIONAL TRANSPORT COSTS FOR LUMBER
1977 ($ PER MBFt., DEFLATED)
NW PSW IL NC NE
5.39 16.02 17.07 26.54 31.21
25.92 25.85 24.25 15.45 15.67
11.61 11. 72 17.68 26.78 31.24
21.25 19.38 8.61 24.18 28.95
TABLE XIII
INTERREGIONAL TR&~SPORT COSTS FOR PLYWOOD
1977 ($ PER MBFt. 2 , DEFLATED)
NW PSW IL NC NE
5.74 16.34 17.20 26.98 31. 72
29.22 29.22 26.60 15.65 16.69
11.81 11.91 17.97 27.23 31.69
22.15 20.25 8.75 24.58 29.43
87
SO
28.54
9.35
27.41
25.48
SO
27.82
9.50
27.79
25.90
88
TABLE XIV
REGIONAL INDEXES OF ACCESSIBILITY LUMBER AND PLYWOOD
Lumber Plywood
Year NW SO SW IL SO NW
1951 169.70 274.49 179.52 154.45 39.84 17.02 1952 169.38 244.31 182.08 162.50 26.82 14.51 1953 168.20 212.20 185.97 166.04 17.15 12.83 1954 183.39 255.95 190.81 177.65 22.88 14.28 1955 206.43 313.48 205.85 196.87 33.32 17.55 1956 205.99 296.47 213.85 196.04 31.82 17.70 1957 197.18 313.36 209.15 189.03 34.50 15.11 1958 178.10 271. 96 193.35 167.53 26.72 12.61 1959 177.48 269.11 188.39 168.56 26.32 12.92 1960 159.25 263.29 166.10 158.42 26.56 10.74 1961 166.65 256.03 182.14 166.73 23.86 10.59 1962 158.52 237.30 166.78 146.74 22.53 11.53 1963 146.83 223.35 160.30 139.73 19.97 9.34 1964 135.00 219.03 144.96 130.63 19.55 8.04 1965 122.43 193.07 129.19 115.42 16.78 7.66 1966 112.71 165.36 107.18 98.45 14.37 8.89 1967 105.17 143.56 100.84 89.89 11.95 8.47 1968 103.77 154.34 97.87 93.03 13.33 7.96 1969 108.92 148.87 106.01 94.54 12.23 8.34 1970 119.25 180.83 117.70 110.72 14.11 7.68 1971 129.88 223.84 132.83 128.79 18.05 6.90 1972 105.78 189.21 110.12 109.54 15.93 5.33 1973 107.42 177.65 110.51 104.94 16.92 6.92 1974 111.99 174.84 111.54 101.49 17.19 8.40 1975 115.18 175.80 115.18 105.03 16.43 8.41 1976 112.01 163.40 112.52 101.43 14.45 8.00 1977 121.08 167.91 119.84 108.20 14.49 8.95 1978 134.02 174.03 131. 25 116.25 14.85 10.52 1979 138.42 188.35 134.05 120.27 16.21 10.70 1980 139.73 193.04 136.83 119.65 15.23 10.36 1981 110.37 165.78 112.15 102.72 17.14 6.77 1982 70.32 140.53 71.01 74.62 17.20 3.18 1983 74.03 134.34 79.99 77.90 19.06 3.07 1984 69.13 131. 06 66.80 71.87 21.40 3.17
For derivation see text.
CHAPTER IV
MODEL VALIDATION
A model is, by definition, a representation of real
world phenomena. A "good" model, it hence follows, is one
that is true to the original. Likewise, the "goodness" of an
econometric model is evaluated by the extent of its accuracy
in depicting the economic system it intends to describe.
Evaluation of an econometric model, as Dhrymes and
others (1972) have pointed out, is a continuous process that
accompanies every step in model specification, estimation,
and implementation. There is a wide array of criteria,
parametric and otherwise, that can be used to evaluate
performance of econometric models. Choice of specific
criteria, and the degree of the stringency with which we
apply each, depend on the stage in model development and the
purpose for which the model is built. In this chapter we
will consider various properties of individual components of
the model developed in Chapter III and evaluate the perform
ance of the model as a whole.
VALIDATION IN PARAMETER ESTIMATES
Construction of an econometric model involves much
more than the working out of a plausible configuration for
sets of seemingly related economic variables. The primary
90
Orcutt iterative procedure in the second stage of
estimation. In case of the latter equation, however, the
estimation method involved precluded the use of such
solution scheme and the original estimates were retained.
With the exception of two market share equations (the South
and the Inland regions), the statistical fit of all
equations, measured in terms of their respective
coefficients of determination, appear acceptable. Of the
total of seventy-two estimated coefficients in the model,
fifty-one are statistically significant at the .01 and
eleven more at the .1 level.
statistical significance of estimated coefficients is
a highly desirable property in an econometric model for it
shows the significance of the information conveyed by
particular data. Yet, statistical significance by itself
does not constitute sufficient ground for validity. The
estimated coefficients must also make sense. A second --and
perhaps a more relevent-- perspective on model validity can
be gained by examining the accuracy of model parameters.
There are, however, no formal procedures for insuring that
the coefficients are of the right magnitudes. The decision
as to what constitutes the right size for a given
coefficient is a judgemental one that depends on the
analyst's knowledge of the market under investigation.
Fortunately, the literature on the wood products
economy is rich with predictions about the magnitude of
certain effects. The more recent econometric studies, in
91
Orcutt iterative procedure in the second stage of
estimation. In case of the latter equation, however, the
estimation method involved precluded the use of such
solution scheme and the original estimates were retained.
with the exception of two market share equations (the South
and the Inland regions), the statistical fit of all
equations, measured in terms of their respective
coefficients of determination, appear acceptable. Of the
total of seventy-two estimated coefficients in the model,
fifty-one are statistically significant at the .01 and
eleven more at the .1 level.
statistical significance of estimated coefficient is a
highly desirable property in an econometric model for it
shows the significance of the information conveyed by
particular data. Yet, statistical significance by itself
does not constitute sufficient ground for validity. The
estimated coefficients must also make sense. A second --and
perhaps a more relevent-- perspective on model validity can
be gained by examining the accuracy of model parameters.
There are, however, no formal procedures for insuring that
the coefficient are of the right magnitudes. The decision as
to what constitutes the right size for a given coefficient
is a judgemental one that depends on the analyst's knowledge
of the market under investigation.
Fortunately, the literature on the wood products
economy is rich wi th predictions about the magni tude of
certain effects. The more recent econometric studies, in
92
particular, provide much insight in this area. An
examination of certain key parameters in the model such as
relative impacts of various exogenous variables and price
elasticities reveals that these estimates are quite compar
able with those obtained in other studies.
The multiregional structure of the present model
imposes a further condition for evaluating the accuracy of
estimated coefficients: that for any particular behavioral
relationship, variations in functional properties across
various regions must be justifiable and reasonably consis
tent with actual market experience in each region. This
condition was employed most intensively in construction of
regional market share relationships, where essentially
identical specifications are used across all regions. This
condition formed the basis for choosing among alternative
formulations of the accessibility index and price terms in
these equations.
cross-regional differences in response to movements in
various exogenous variables can be analyzed by examining the
relevant impact multipliers. Impact multipliers for regional
lumber outputs associated with selected explanatory
variables are reported in Table XV.
These multipliers are computed numerically by
successive sUbstitution of relevant equations in the model.
Multipliers associated with other exogenous variables not
reported in Table XV, can also be obtained in the same
fashion. Due to the rather complex structure of the access-
93
ibi1ity index, however, a simple analytical solution for the
impact multipliers associated with the Producer Price Index
for Lumber and Wood Products Freight is difficult to obtain.
The reported multipliers in this case were computed
numerically at the means for each variable. Figures in each
column of Table XV show the effects on regional lumber
outputs that is likely to result from a unit increase in
variables listed in the first column.
TABLE XV
IMPACT MULTIPLIERS FOR SELECTED EXPLANATORY VARIABLES: REGIONAL LUMBER OUTPUT
(Measured as % of Total Consumption)
NW SO SW IL Housing starts •••• : 0.0035 0.0024 0.0018 0.0011
(xl000) Personal income ••• : 0.0174 0.0132 0.0085 0.0044
(MMM 1972 $) Average lumber price in U.S •••••• : 0.218 0.152 0.149 0.067
(1967 $ / MBFt.) Regional lumber prices •••••••••••• : -0.218 -0.152 -0.149 -0.067
(1967 $ / MBFt.) Regional stumpage prices •••••••••••• : -0.120 -0.110 -0.119 -0.085
(1967 $ / MBFt.) PPI wood products freight ........... : -0.0035 -0.0014 -0.0013 -0.0011
(1967 = 100)
CAN 0.0059
0.0237
0.1090
0.0160
These multipliers are also the mechanisms by which
changes in aggregate consumption can be allocated across
producing regions. This can be illustrated by an example.
Suppose there is an increase of one thousand units in
national housing starts. From equation (1), Table VIII, we
94
would expect a corresponding increase, on the average, of
approximately 5.1 MMBFt. in total lumber consumption. The
multipliers in the first row of Table XV, can now be used to
determine how such increase in aggregate demand will be
allocated across producing regions. The results show that on
the average, holding all other variables constant, such an
increase in national housing starts would lead to increases
of 1.2, 0.9, 0.61, 0.47, and 1.98 MMBFt. respectively in the
NW, SO, SW, IL, and Canadian imports. Due to errors in
coefficient estimates, the total does not add to 5.1 MMBFt.
The discrepency, however, is negligible and can be corrected
by proportional adjustment of multipliers. This procedure
can b~ appli~d to determine the relative impacts on regional
market shares of unit change in these variables. Regional
market share impacts for selected variables are reported in
Table XVI.
The dynamic context of the present model and inter
dependencies that are built into it, limit the usefulness of
analysing multipliers in a static setting. These multipliers
do, however convey important descriptive information on
parameter estimates and behavioral consistency of individual
relationships. Examination of impact multipliers in Tables
XV and XVI, for example, reveals some interesting patterns
of differential response to various exogenous factors across
the four domestic supply regions.
It is apparent from these results that short-run
impacts on market shares of the variables considered here
95
TABLE XVI
IMPACT MULTIPLIERS FOR SELECTED EXPLANATORY VARIABLES: REGIONAL JGUMBER MARKET SHARES
NW SO SW IL CAN Housing starts.: .00059 -.00051 -.00116 -.00176 .00294
(x1000) Personal income: .0039 -.00018 -.0050 -.0091 .0100
(MMM 1972 $) Average lumber price in u.S ••• : .0790 .0130 .0100 -.0720 -.0300
( • 67 $ /MBFt.) Regional lumber prices ••••••••• : -.2180 -.1520 -.1490 -.0670 -.0160
( • 67 $ /MBFt.) Regional stump. prices ••••••••• : -.1200 -.1100 -.1190 -.0850
( • 67 $ /MBFt.) PPI Wood Prod. freight •••••••• : -.00167 .00043 .00052 .00072
(1967 = 100)
are generally quite small. The NW, however, appears to be
relatively more sensitive to these exogenous impacts than
other regions. Market shares for the NW appear to be
particularly sensitive to transport costs. In the case of
the SW region, the magnitude of some impacts --particularly
those associated with transport costs and price variables-
are somewhat higher than expected. For a net importing
region such as the SW, it is indeed reasonable to expect the
market share to be inversely related to accessibility. Such
expectation is justified because a decline in transport
costs (i.e. increased accessibility) can in effect make the
region itself more accessible to other producing regions and
invite competition.
96
A possible explanation for the observed sensitivity of
the SW market shares to accessibility lies in the species
composition of the region's lumber products. Redwood, which
constitutes a large portion of the lumber produced in the SW
region, is a specialty product in demand in all other
regions in the country. Thus, though a net importer of
lumber products; the region has historically exported
considerable volumes of redwood to other regions.
As indicated in Chapter I, the coefficients on price
terms in market share relationships measure the degree of
competitiveness and reflect the extent of regional
substitution in the product market. If products from
competing regions are perfect substitutes, then small
changes in relative prices in one region can be expected to
result in large changes in the demand facing that region. As
evidenced by the low elasticities on price terms (ranging
between 0.02 for the NW and 0.1 for the SW), this clearly is
not the case in the lumber market. The low price responsive
ness of regional market shares found here suggest that the
extent of regional sUbstitution in the lumber market is
indeed quite limited.
In view of the low price elasticities of regional
lumber supplies (ranging from 0.08 for the NW and 0.15 for
the SO), this finding is not all too surprising. A more
serious implication here is that this finding also casts
doubt on the validity of the product homogeneity assumption.
There is no doubt that consumers' species preferences has
97
some influence on determination of demand for lumber
produced in specific regions. The extent of this influence,
however, cannot be deduced directly within the context of
the present model.
Price elasticities of demand for regional plywood, on
the other hand, are significantly greater than those found
for lumber, indicating a much higher level of competitive
ness in this market. Estimated elasticities indicate that a
one percent change in plywood price ratios can be expected
to lead to 1.3 and 2.0 percent change, respectively, in
demand for plywood produced in the NW and the so.
V~DATION IN SIMULATION AND FORECASTING
Validity in simulation performance was used as the
primary guideline in evaluation of the model. As indicated
earlier, the objective in constructing the present model
was, in part, simulation analysis and forecasting. An
important test of the model's validi ty is, therefore, its
efficacy in serving these ends. To test the adequacy in
simulation performance of the model, we will first consider
the accuracy with which the model can replicate the actual
time paths of endogenous variables through a graphical
analysis of historical and simulated series; and then use
the model to develop proj ections of key variables for two
years beyond the estimation period.
A comparisons of actual and predicted series for all
enodogenous variables are shown in Figures 10 through 30.
~ • " ILC
III 0 l· l ~ t
.. 4J
42
41
40
311
J8
37
38
35
34
33
J2
31
JO
29
25
98
51 53 55 57 511 81 83 55 87 811 71 73 75 n 711 81 83
- Aclual + P...s1cI..t
Figure 10. Historical simulation total lumber consumption.
lJO~----------------------------~--------------~
120
110
100
90
80
51 53 55 57 511 8' 8l as 87 eg 71 73 75 77 711 8' IIJ
Figure 11. Historical simulation average lumber price.
99
21
20
19
111
17
18
15
14
13
12
11
10
9
S
7
8
5
4
3
2
51 53 55 57 58 81 a3 as 87 ee 71 73 75 77 78 81 83
Figure 12. Historical simulation total plywood consumption.
8 I ~ ii • " .5
~~-----------------------------------------------,
320
300
280
260
240
220
200
1110
160
140
120
51 53 55 57 58 81 a3 as 87 88 71 73 7!l n 79 111 a3
- Actual + Pr.dIc:tM
Figure 13. Historical simulation PPI softwood plywood.
... m
eo.-------------------------------------~------_,
55
45
:1 .a " .. " .. 35 (I
30
25
20
- Actual + PNdcbod
Figure 14. Historical simulation stumpage price NW.
... m :1
~ rCI (I
2a,-------------------------------------------~~
27
25
25
24
23
22
21
20
10
15
17 iii I I • I • iii.
51 53 55 57 50 51 53 55 51 ell 11 13 75 n 711 III 53
- Actual
100
Figure 15. Historical simulation stumpage price so.
~~--------------------------------------~-------~
45
40
25
20
51 53 55 51 5. 81 113 55 81 811 11 13 75 n 79 81 113
- Actual + P...-.:ted
Figure 16. Historical simulation stumpage price sw .
It.. III :I ".. " .. at
.#,'
~,------------------------------~------------~
28
28
24
22
20
111 i 111 i 14
12
10
II
:1~-r,,1-r"1-r"1-r"'-r, ~,~,~,~'~,~'-r,~'-r,~'-r,~'-r,~'-,rT'-'~'~I~'~I~' 51 53 55 57 511 81 113 115 117 1111 11 73 75 71 79 81 113
- Actual
101
Figure 17. Historical simulation stumpage price IL.
102
140
130
120
110 ... CD :2 "-• 100 .... .. ~
~
51 53 55 51 51 11 4J 85 111 811 11 13 15 n 79 111 III
- Actual
Figure 18. Historical simulation lumber price NW.
150
Al 140
130 I
... CD 120 :2 "-• .... ..
110 ~
51 53 55 51 5' 111 III 115 81 III 11 13 75 71 7' 111 III
- Actuat
Figure 19. Historical simulation lumber price so.
103
leo~----------------------------------------,
150
140
lJO "-III Ji "-.. 120 ... .. ~
110
100
, ' , ' , ' , ' , ' , ' , ' , ' , ' , ' , ' , ' , ' , ' , ' , I 110 51 53 55 57 51 81 III 85 87 81 71 73 75 77 7a 81 III
- Actual + Predicted
Figure 20. Historical simulation lumber price sw.
lJO
120
110
"-III :J "-.. 100 ... .. ~
"~ :l~,
51 53 55 57 sa 81 III 85 87 sa 71 73 75 77 7a '" III
- Actual
Figure 21. Historical simulation lumber price IL.
8 I
" " at
• • " E
8 I
" " ~ x • " S
'~.r--------------------------------~----------~
140
130
120
110
100
140
130
120
110
100
90
eo~ I
70 J I
1501 51
Actual
Figure 22. Plywood price Douglas fir.
57 511
Actual
f\, \ / \
I
'/
81 83
+ Pro6:I ...
85 87 811
Figure 23. Plywood price Southern pine.
104
c 2 a. E , • c o o a ~
32 -----------_. __ .
51 53 55 57 5a 111 III 115 87 ea 71 73 75 n 7a 111 III
- Actual
Figure 24. Lumber market share Canada.
RIIS EtTo ... l.35 3a,------------------------------------------------~
c
J8
37
J8
35
l4
o Jl ::: e J2 , ~ 31 o o JO
29
21S
27
211
25 1 24 -l
I
23 -l 22 I
~\ ! ~ I
.,.,.,.,.,.,.,.,.,.,.,.,.,~ 51 53 55 57 5a 111 III e 117 sa 71 73 75 77 7a lSI III
- Actual + Pred1ctod
Figure 25. Lumber market share NW.
105
c o :> Go E , • c o U
o ~ It
c ~ D-E , • c 0 U
0 ~ It
106
NIlS Emw-o.71 29~----------------------------------------------,
21!
27
28
24
23
22 21
20
10 ~ fl. 1'1 iii
51 53 55 57 58 81 83 85 87 69 71 73 75 77 79 I!I 1!3
- Adual
Figure 26. Lumber market share so.
MIS Emw-o.71 HI
II!
17
11!
51 53 55 57 511 81 1!3 as 87 811 71 73 75 n 711 81 1!3
- Actual
Figure 27. Lumber market share SW.
c o ., Do E , • c a (J
a o 0-
It
c a ., Do E , . c a (J
a ~ It
:: ~ 121 11
10
II
II
Sl !SJ SS 57 511 81 113 lIS 87 8e 71 73 75 n 78 111 113
- Actual + Preclc:ted
Figure 28. Lumber market share IL.
RIolS Erro.-2.2 115
110
~ 7S
70 \ , ~ 55
80
S5
:L 51 !SJ SS S7 511 81 III lIS 87 811
Actual + ""-dIeted
Figure 29. Plywood market share NW.
107
71
RMS ErTor-2.6 60~--------------------------------------------~---,
:;0
~ ~ .. Q.
E , . ~ 30 u "0
~ I( 20
10
51
10
II
II
7 c a .. Q. II E , . c 5 a 0
"0
~ 4
Ie 3
2
53 55 57 59 111 153 115 117 69 71
Actual + PredIcted
Figure 30. Plywood market share so.
RI.IS Enot - O.gS
+ ProodIcted
Figure 31. Lumber market share North
108
109
For the seven key variables, i.e. regional lumber and
plywood market shares, "root-mean-square" (rms) simulation
errors are also reported.
In general, the results of ex post simulations appear
to support the overall structure of the model. Nearly all
simulated series seem to track the principal trends closely
and, in most cases, replicate the timing of turning points
accurately. In the aggregate market, the results for total
product consumption are superior to product price outcomes
for both lumber and plywood. simulations of regional
stumpage and product prices are of mixed quality. Although
in all cases long-run trends are closely approximated, in
several instances such as stumpage prices in the NW and so
short-run fluctuations and the timing of several peaks and
troughs are poorly depicted.
In evaluating the simulation performance of the model,
the heaviest weight was given to regional market share
variables. Because of their pivotal role, and the fact that
market shares consti tute the primary output of the model,
the ultimate validity of the entire model is decided by the
performance of these relationships. In judging the accuracy
of market share predictions, it is important to recognize
the serious implications of bias in terms of actual quanti
ties. A seemingly negligible error in these variables, once
translated into product quantities, can lead to unacceptable
results in estimates of regional outputs.
110
Historical simulations for regional market shares are
shown in Figures 25 through 30. A graphical comparison of
these results, shows that in all cases historical trends are
reproduced remarkably well. The relatively low rms errors in
these simulations also verify this conclusion. However, in a
model designed for short-run projections, it is also
essential that short term fluctuations be approximated
accurately. As evidenced by the simulation results, it is
apparent that in some cases this condition is not satisfied
as fully as we would have liked.
In general, the results obtained for Canada, NW, and
sw appear to be much more representative of the historical
behavior than those found for other regions. For the last
several years of the sample period, South's shares of lumber
and plywood markets are both under estimated. In the case of
the IL lumber market shares, many short term fluctuations
are either missed or represented with delay.
The untimely simulation of short term peaks and
troughs of historical series appears as a common problem in
several market share relationships. This problem is, in
part, attributable to the autoregressive structures of
market share relationships. In all cases, however,
deviations from actual values are very small in absolute
magnitude and can be corrected by making minor adjustments
in few parameters of the model.
The historical trend in the share of the North region
in the lumber market is shown in Figure 31. Since the region
111
was not included in the model, the predicted values were
derived by subtracting cumulative market shares of other
regions in each period from 100 percent. It is important to
note that since market shares for this region are determined
exogenously, discrepencies between the predicted and actual
values represent the magnitude of total error in market
share equations. comparison of actual and simulated series,
as shown in Figure 31, show that the cumulate error in
market share estimates is indeed negligible.
As a final check on the performance of the model as a
whole, the complete model will be simulated forward for two
years beyond the estimation period. This ex post forecasting
exercise has the two-fold purpose of showing the predictive
accuracy of the model, and providing a means for judging the
overall utility of the model as a forecasting tool.
Values for nearly all exogenous variables for both
years are available from same sources as listed in Appendix
A. Also available, are the 1985 values for certain key
endogenous variables such as aggregate consumption and
regional outputs. Thus it is posssible to compare model
projections with known values of certain variables. S:tnce
1986 values for some exogenous variables were not available,
it was necessary to use estimated values derived under
certain assumptions. For instance the national forest
offerings were held constant at the 1985 level. Change in
output per employee for all regions were assumed to continue
112
TABLE XVII
PROJECTED VALUES FOR MAJOR SOFTWOOD LUMBER AND PLYWOOD MARKET VARIABLES
Variable 1984 1985 1986 Total Lumber Cocsumption •• : 43021 44207.4*
(44204) 45867.3
Average Price for lumber ••• : 81. 79 79.82 88.17
Total plywood consumption •• : 19508 22735 23618.3
(22838) PPI softwood plywood •••••• : 303.5 307.2 314.9
(302.9) Average lumber price NW ••.•• : 77.70 73.35 84.90
Average lumber price SO ••••• : 99.40 104.7 109.0
Average lumber price SW ••••• : 110.29 106.2 114.9
Average lumber price IL ••••• : 81.40 75.13 74.70
Plywood production NW: 8319.2 8281. 9 8319.2
(8395.5) Plywood production SO: 10351.7 11406.4 11912.6
(11033.8) Mkt. share lumber Canada: 30.80 32.50 33.48
(33.04) Mkt. share lumber NW •••• : 23.41 24.45 24.20
(23.21) Mkt share lumber SO •••• : 24.75 25.14 25.02
(24.54) Mkt. share lumber SW •••• : 8.86 9.30 8.71
(8.79) Mkt. share lumber IL •••• : 9.37 10.39 10.50
(10.32)
* Figures in parantheses show actual values.
113
at their average historical rates in all regions; and the
price of Canadian lumber was estimated at a level commen
surate with the Producer Price Index for all softwood
lumber.
The two-year projections for aggregate levels of
consumption and prices for both lumber and plywood, and
regional product prices and market shares are shown in Table
XVII. Plywood price index and regional product and stumpage
prices were all estimated from reduced form equations so
that current endogenous variables were eliminated from the
right-hand side of these equations.
A comparison of proj ected and actual values of the
main market variables in 1985 indicates that nearly in all
cases the actual values are closely approximated. The major
exception is the regional plywood market. Plywood production
in the NW is slightly underestimated, while production in
the SO is overestimated by nearly 400 million square feet
(1.6 percent of total consumption). PPI for softwood plywood
also appears to miss the 1985 mark by six points.
During this two year period, the projected trend is
one of increasing demand in both lumber and plywood markets
stimulated by the surge in the national housing market. with
the exception of the southern pine lumber, 1985 product
prices are projected to fall below their 1984 levels and
then increase slightly in 1986.
projected pattern for regional market shares in lumber
show that increases in quanti ties of lumber consumed will
114
not be distributed evenly across producing regions. As
indicated by resul ts in Table XVII, domestic producers'
shares of the lumber market are expected to remain largely
stable. Canadian producers appear as the single most
important beneficiary of the increased demand for lumber.
The market share outcomes suggest that nearly all of the
anticipated increments in demand for lumber will be
satisfied by Canadian imports; thus adding to the prominence
of Canada as the major supplier of lumber in the u.s.
market.
Canada's role in the domestic lumber market is,
however, sensitive to price conditions both in Canada and in
the u. S. Shifts in the delivered prices of the Canadian
lumber are likely to take place as a result of changes in
exchange rates and/or imposition of tariffs. In view of
recent trends in the international currency market, and the
current debate over the imposition of tariff on the Canadian
lumber imports, the position of Canada in the u.S. lumber
market is likely to undergo some change in the future.
The results of the present study indicate, for
example, that imposition of a 20 percent ad valorem duty on
lumber imports can result in a 0.75 percent (350MM BFt.)
reduction in Canada's share of the lumber market in the
first year. Such impact, however, is not likely to persist
in future periods. The magnitudes of future period impacts
also depend on price conditions in the u.S. For one thing,
any such change in the delivered price of the Canadian
115
lumber will also lead to increased average prices in the
u.s. (and possibly some reduction in quantities demanded).
We can therefore expect the initial period price disad
vantage arising from the imposition of tariffs to be some
what mitigated in the following periods so that the overall
impact on market shares will not be very large.
CONCLUDING REMARKS
This research was motivated by an interest in
exploring the mechanisms that underlie the geographical
distribution of production in the lumber and wood products
industry as a specific case in the broader context of
spatial competition. The research is a further step in the
ongoing efforts in regional modeling, and provides a
framework for the explicit treatment of linkages between the
national and regional economies. In addressing this problem,
we opted for an econometric approach as a compromise between
the oversimplicity of some conventional methods of location
analysis and the complexity of mathematical programming.
The design of the model was guided by the three-fold
objective of developing a model that accurately describes
the behavior of the two industries and their spatial
characteristics, can provide reliable projections of short
term trends, and serve as a tool in simulation experiments.
The overall performance of the model bears witness to
the fact that all three objectives were met, though with
varying degrees of success. with respect to the approach
116
adopted here, it has been shown that it is indeed a viable
one. The findings provide ample evidence in support of the
initial postulates of the study regarding the determinants
of regional market shares.
In several areas, however, there is room for some
refinements that can improve the effectiveness of the
approach. There is, first of all, the specification of
spatial units that can be better presented by further
disaggregation. For example, more homogeneous spatial units
can be created by separate treatment of the Coastal and
Inland parts of the Northwest region; and independent
presentation of the Southeast and Southcentral regions. with
further experimentation, and using more appropriate measures
of product demand such as housing starts, alternative
accessibility indexes can be constructed to better represent
regional locational advatage.
Since a primary point of interest in this undertaking
was development of the approach itself and testing of its
validity, much emphasis was placed on the statistical
properties of the model. This, combined with efforts to
maintain theoretical consistency in behavioral relation
ships, led to certain rigidities in the structure that
complicate the forecasting and simulation processes. A
further problem in forecasting application arises from the
frequent appearance of lagged endogenous variables in the
model which can lead to efficiency problems in extended
forecasts.
117
Further extensions along the product dimension of the
model are also possible. The pulpwood sector might, for
instance, be included in the model so that a more complete
picture of competition in regional stumpage markets is
obtained. These and other extensions of the model are,
however, contingent upon availability of data at the
regional level, a persistent problem in all regional
modeling efforts including the present one.
LITERATURE CITED
Adams, D.M. (1976), "The Response of Softwood Stumpage and Product Prices to Changes in National Forest Timber Harvest Schedules". Unpublished report, Pacific Northwest Forest and Range Experiment Station, Portland, Or., 13 pp.
Adams, D.M. (1977), Effects of National Forest Timber Harvest on Softwood Stumpage, Lumber, and Plywood Markets. School of Forestry, Oregon State University, corvallis, Or., Res. Bulletin No. 15.
Adams, D.M., R.W. Haynes, T.J. Mills, D. Shearer, and S. Childress (1979), Production, Consumption, and Prices of Softwood Products in North America, Regional Time Series Data. School of Forestry, Oregon State Univ. Corvallis, Or., Res. Bulletin No. 27.
Adams, D.M. and R.W. Haynes (1980), The 1980 Softwood Timber Assessment Market Model: Structure, projections, and Policy Simulations. Forest Science Monograph no. 22, 64 pp.
Adams, D.M. and R.W. Haynes (1980a), "U.S.-Canadian Lumber Trade: The Effect of Restrictions". in R.A. Sedjo (ed.), Issues in U.S. International Forest Products Trade. Resources For The Future, Research Paper R-23, Washington D.C., 1980.
Adams, D.M., R.W. Haynes, G.F. Dutrow, R.L. Barber, and J.M. Vasievich (1982), "Private Investment in Forest Management and the Long-Term Supply of Timber". Amer. Journal of Agr. Econ. 64, NO.2, pp. 232-41.
Adams, D.M. and R.W. Haynes (1986); "A Spatial Equilibrium Model of U.S. Forest Products Markets for Long-Range Projection and Policy Analysis". TIMS studies in the Management Sciences, North-Holland, pp. 73-87.
Adams, F.G. and J. Blackwell (1973) "An Econometric Model of the united states Forest Products Industry". Forest Science, 19, pp~ 82-97.
Adams, F.G. and N. Glickman (1980), (eds.) Modeling the Multiregional Economic System. Lexington Books, Lex., Mass., pp.147-60.
Alperovich, G. (1980) "Regional Elasticities of Substitution". Journal of Regional Science, 20, No.4, pp. 503-11.
American Plywood Association, "Structural Panel Production by State". Bulletin FA-235, [annual].
Armington, P.S. (1969) "A Theory of Demand for Products Distinguished by Place of Production". International Monetary Fund Staff Papers, 16, pp. 159-76.
119
Ballard, K.P., N.J. Glickman, and R.D. Gustely (1980) "A Bottom-Up Approach to Multiregional Modeling: NRIES". in Adams, F.G. and N.J. Glickman, (eds.) Modeling the Multiregional Economic System. op.cit., pp. 147-69.
Connaghton, K.P. and R.W. Haynes (1983) "An Evaluation of Three Simplified Approaches to Modeling the Regional Demand for National Forest Stumpage". Forest Science, 29, No.2, pp. 3-12.
Deacon, R.T. and M.B. Johnson (1985) (eds.) Forestlands, Public and Private. Pacific Institute for Public Policy Research, Ballinger Pub. Co., Cambridge, Mass.
Dhrymes, P.J., E.P. Howrey, S.H. Hymans, J. Kmenta, E.E. Leamer, R.E. Quandt, J.B. Ramsey, H.T. Shapiro, and V. Karnowitz (1972) "criteria for Evaluation of Econometric Models". Annals of Economic and Social Measurement, 1, No.3, pp. 291-323.
Dorfman, R. (1953) " Mathematical or 'Linear' Programming: A Nonmathematical Exposition", in E. Mansfield (ed.), Microeconomics, Selected Readings. Norton, N.Y., 1979, pp. 137-65.
Dowdle, B. and S.H. Henke (1985) "Public Timber Supply and the Wood-Products Industry", in Deacon, R.T. and M.B. Johnson (eds.), Forestlands, Public and Private. op.cit. pp. 77-102.
Engle, R.F. (1979a) "The Regional Response to Factor Supplies: Estimates for the Boston SMSA", in Wheaton, W.C., (ed.) Interregional Movements and Regional Growth. The Urban Institute, Washington D.C., 1979, pp.157-83.
Engle, R.F. (1979b) "Estimation of Price Elasticity of Demand Facing Metropolitan Producers". Journal of Urban Economics, 6, pp.42-64.
120
Enke, S. (1951) "Equilibrium Among Spatially Separated Markets: Solution by Electric Analogue". Econometrica, 21, PP. 547-66. Reprinted in Dean, R.D., et.al. Spatial Economic Theory. The Free Press, N.Y. 1970, PP. 277-84.
Fehd, C.S. (1975) "Introducing Price Indexes for Railroad Freight". Monthly Labor Rev., (June 1975), pp. 19-24.
Fisher, F.M. (1965) "The Choice of Instrumental Variables in the Estimation of Economy-Wide Econometric Models". in H.M. Blalock Jr., Ed., Causal Models in the Social Sciences. Adline, Chicago, 1971, pp. 245-72.
Fox, K.A. (1953) "A Spatial Equilibrium Model of the Livestock Feed Economy". Econometrica, 21, pp. 547-66.
Garnick, D.H. (1980) "The Regional Statistics System". in Adams, F.G. and N.J. Glickman, (eds.) Modeling the Multiregional Economic System. op.cit., pp.25-38.
Ginsberg, A.L. and R.M. Stern (1965) "The Determination of Factors Affecting American and British Exports in the Inter-War and Post-War Periods". Oxford Economic Papers, 17, pp. 263-78.
Ginsburgh, V.A. and J.L. Waelbroeck (1981) Activity Analysis and General Equilibrium Modeling. North-Holland, Amsterdam.
Gregory, G.R. (1960) "A statistical Investigation Of Factors Affecting the Market for Hardwood Flooring". Forest Science, 6, pp. 123-34.
Gregory, G.R., D. Hair, H.R. Josephson, V. Holopainen, K. Mantel, S.L. Pringle, P. Riihinen, R. Saether, and H.J. Vaux (1971) "Forecasting in Forestry and Timber Economy". Folia Forestalia, 101, pp. 1-49.
Haynes, R.W. (1977) "A Derived Demand Approach to Estimating the Linkage Between Stumpage and Lumber Markets". Forest Science, 23, No.2, pp. 281-8).
Haynes, R.W., D.L. Holley, and R.A. King (1978) A Recursive Spatial Equilibrium Model of the Softwood Timber Sector. N.C. State Univ., School of Forest Resources, Technical Report No. 57.
Haynes, R.W., K.P. connaghton, and D.M. Adams (1981) Projections of the Demand for National Forest Stumpage by Region; 1980-2030. USDA/FS, Research Paper, PNW-282.
121
Henderson, J.M. and R.E. Quandt (1980) Microeconomic Theory, A Mathematical A~proach. McGraw-Hill, N.Y.
Holland, I.I. (1960) "An Explanation of Changing Lumber Consumption and Price" Forest Science, 6, No.2, pp. 112-24.
Holley, D.L., R.W. Haynes and H.F. Kaiser (1975) An Interregional Timber Model for Simulating Changes in the Softwood Timber Economy. N.C. State Univ. School of Forest Resources, Technical Report No. 54.
Houston, D.B. (1967) liThe Shift and Share Analysis of Regional Growth: A critique". Southern Econ. Journal, 33, (April 1967), PP. 577-82.
Howe, C.W. (1979) Natural Resource Economics: Issues, Analysis, and policy. J. Wiley & Sons, N.Y.
Humphrey, D.B. and J.R. Moroney (1975), "Substitution Among Capital, Labor, and Natural Resource Products in American Manufacturing". Journal of Political Economy, 83, pp. 57-82.
Hyde, W.F. (1980) Timber supply, Land Allocation, and Economic Efficiency. Resources for the Future, Wash. D.C., John Hopkins Univ. Press, Baltimore, Md.
Intriligator, M.D. (1971) Mathematical optimization and Economic Theory. Prentice Hall, N.J.
Isard, W. (1960) Methods of Regional Analysis. MIT Press, Cambridge, Mass.
Johansson, P.o. and K.G. Lofgren (1985) The Economics of Forestry and Natural Resources. Basil Blackwell, Oxford.
Johnston, J. (1984) Econometric Methods. 3rd. ed., McGrawHill, N.Y.
Kennedy, M. (1974) "An Economic Model of the World oil Market". The Bell Journal of Economics, 5, pp. 540-77.
Kloek, T. and L.B.N. Mennes (1960) "Simultaneous Equation Estimation Based on Principal Components of Predetermined Variables". Econometrica, 28, pp. 45-61.
Labys, W.C. (1973) Dynamic Commodity Models: Specification, Estimation. and Simulation. Lexington Books, Lex., Mass.
Leamer, E.E., and R.M. stern (1970) Quantitative International Economics. Allyn and Bacon, Boston.
Lindell, G.R. (1979) "World Softwood Lumber Trade: Patterns, Trends, and Prospects". Forest Products Journal, 29, No.7, pp. 43-9.
McCarl, B.A. and R.W. Haynes (1985) "Exchange Rates Influence Softwood Lumber Trade". Journal of Forestry, 83, No.6, pp. 368-70.
122
McCarthy, M.D. (1971) "Notes on the Selection of Instruments for Two stage Least Squares and K. Class Type Estimators of Large Models". Southern Econ. Journal, 37, pp. 251-8.
McKillop, W.L. (1967) "Supply and Demand for Forest Products: An Econometric Study". Hilgardia, 38, Univ. of Calif. Agr. Exp. Station, pp. 1-132.
McKillop, W.L. (1969) "An Econometric Model of the Market for Redwood Lumber", Forest Science, 15, pp. 159-70.
Merrifield, D.E. and R.W. Haynes (1983) "Production Function Analysis and Market Adjustments: An Application to the Pacific Northwest Forest Products Industry". Forest Science, 29, No.4, pp. 813-22.
Merrifield, D.E. and R.W. Haynes (1984) "The Adjustment of Product and Factor Markets: An Application to the Pacific Northwest Forest products Industry". American Journal of Agricultural Economics, 66, pp. 79-87.
Mills, T.J. and R.S. Manthy (1974) An Econometric Analysis of Factors Determining supply and Demand for Softwood Lumber. Michigan state Univ., Agricultural Experiment station, East Lansing, Research Report No. 238, 59 pp.
Milne, W.J., F.G. Adams, and N.J. Glickman, "A Top-Down Multiregional Model of the U.S. Economy". in Adams, F.G. and N.J. Glickman, (eds.) Modeling the Multiregional Economic system. op.cit., pp.133-45.
Nautiyal, J.C. and B.K. Singh (1985) "Production Structure and Derived Demand for Factor Inputs in the Canadian Lumber Industry". Forest science, 31, No.4, pp. 871-81.
Nelson, R.H. (1985) "Mythology Instead of Analysis, the story of Public Forest Management". in Deacon, R.T. and M.B. Johnson (eds.), Forestlands. Public and Private. Ope cit. pp. 23-76.
123
Nerlove, M. (1958a) "Distributed Lags and Estimation of Long-Run Supply and Demand Elasticities: Theoretical Considerations". Journal of Farm Economics, 40, NO.2, pp. 301-15.
Nerlove, M. (1958b) "statistical Estimation of Long-Run Elasticities of Supply and Demand". Journal of Farm Economics, 40, No.4, pp. 861-81.
Niedercorn, J.H. and B.V. Bechdolt (1969) "An Economic Derivation of the 'Gravity Law' of spatial Interaction". Journal of Regional Science, 10, pp. 273-82.
Pindyck, R.S. and D.L. Rubinfeld (1981) Econometric Models and Economic Forecasts. 2nd. ed., McGraw-Hill, N.Y.
Polak, J.J. (1950) "Notes on the Measurement of Elasticity of Substitution in International Trade". Rev. of Economics and Statistics, 32, No.1, pp 16-20.
Pringle, S.L. (1971) "Substitution as a Problem in Forecasting". in Gregory, et ale (1971) op. ci t.
Richardson, H.W. (1979) Regional Economics., Univ. of Ill. Press, Chicago.
Robinson, V.L. (1974) "An Econometric Model of Softwood Lumber and Stumpage Markets, 1949-1976". Forest Science, 20, pp. 171-9.
Robinson, V.L. (1975) "An Estimate of Technological Change in the Lumber and Wood-Products Industry". Forest Science, 21, pp. 149=54.
Samuelson, P.A. (1952) "Spatial Price Equilibrium and Linear programming". American Econ. Rev., 42, pp. 283-303.
Samuelson, P.A. (1976) "Economics of Forestry in an Evolving Society". Economic Inquiry, 14, (Dec. 1976), pp.466-91.
Schallau, C.H. and W.R. Maki (1986) Economic Impacts of Interregional Competition in the Forest Products Industry During the 1970's. USDA/Forest Service, Pacific Northwest Research Station, Research Paper PNW-350, 43 pp.
Spelter, H. and R.G. Anderson (1985) A Profile of Wood Use in Nonresidential Building Construction. USDA/Forest Service, Resource Bulletin FPL-15, 22 pp.
stanford Research Institute (1954) America's Demand for Wood. A Report to Weyerhaeuser Company, Tacoma, Wa.
124
Stier, J.C. (1980) "Estimating the Production Technology in the u.s. Forest Products Industries". Forest Science, 26, No.3, pp. 471-82.
Stigler, G. J. (1968) The organization of Industry. Univ. of Chicago Press, Chicago, Ill.
Takayama, T. and G.G. Judge (1964a) "Equilibrium Among Spatially Separated Markets: A Reformulation". Econometrica, 32, pp. 510-24.
Takayama, T. and G.G. Judge (1964b) "Spatial Equilibrium and Quadratic Programming". Journal of Farm Econ., 44, pp. 67-93.
Takayama, T. and G.G. Judge (1971) spatial and Temporal Price and Allocation Models. North-Holland, Amsterdam.
Tinbergen, J. (1946) "Some Measurements of Elasticities of Substitution". Rev. of Economics and Statistics, 28, No.3, pp. 109-16.
Tomek, W.G. and K.L. Robinson (1952) Agricultural Product Prices. Cornell Univ. Press, N.Y.
USDA/Forest Service (1973) The Outlook for Timber in the united states. Forest Resource Report No. 20, 377 pp.
USDA/Forest Service (1982) An Analysis of the Timber situation in the United states 1952-2030. Forest Resource Report No. 23, 499 pp.
Uzawa, H. (1959) "Prices of the Factors of Production in International Trade". Economica, 27, pp. 448-68.
Western Wood Products Association, Statistical Yearbook of the Western Lumber Industry. (annual issues), Portland, Oregon.
Wilson, A.G. (1967) "A statistical Theory of Spatial Distribution". Transportation Research, 1, pp. 253-68.
Zellner, A. (1962) "An Efficient Method of Estimating Seemingly Unrelated Regressions and Tests for Aggregation Bias". Journal of the American statistical Association, 57, pp. 348-68.
Zellner, A. and H. Theil (1962) "Three-Stage-Least-Squares: Simultaneous Estimation of Simultaneous Equations". Econometrica, 30, pp. 54-78.
APPENDIX A
DATA SOURCES
1 - Adams, D.M., K.C. Jackson, and R.W. Haynes (1986) "Production, Consumption and Prices of Softwood Products in North America: Regional Time Series Data, 1950-1985". Unpublished Report, USDA/FS, Pacific Northwest Forest and Range Experiment station, Portland, Or.
125
2 - USDA-Forest service, Pacific Northwest Research Station, Production, Prices, Employment, and Trade in the Northwest Forest Industries. Resource Bulletin, PNW-13 0 , [quarterly issues]. Portland, Ore.
3 - National Forest Products Association, Fingertip Facts and Figures. [monthly], Washington, D.C.
4 - USDC, Business and Defense Service Admin. U.s. Lumber Exports: 1953-1966. 1968.
5 - USDC, Bureau of the Census, County Business Patterns. [annual issues].
6 - USDC, Bureau of the Census, "Lumber Production and Mill Stocks", Current Industrial Reports, Series MA-24T [annual].
7 - American Plywood Association, "structural Panel Produuction by States". Bulletin FA-235, [annual], Tacoma, Wa.
8 - Economic Report of the President. U.s. Gov. Printing Office, 1985.
9 - USDL, Bureau of Labor Statistics, Employment and Earnings. [monthly].
10- U.s. Federal Reserve System, Federal Reserve Bulletin. [monthly].
11- USDC, Bureau of the Census, "Value of New Construction Put in Place in the U.s.", Construction Report Series C30-80S.
12- USDL, Bureau of Labor statistics, Employment and
126
Earnings, states and Areas, 1939-1978., and Supplements.
13- USDA, Forest Service, U.S. Timber Production, Trade, Consumption. and Price statistics, 1950-84. Miscellaneous Publication No. 1442.
14- USDL, Bureau of Labor Statistics, Producer Price and Price Indexes. [monthly and annual].
15- USDC, Bureau of the Census, statistical Abstracts of the united States. [annual].
16- USDL, Productivity Measures for Selected Industries. Bulletin No. 2256.
17- USDC, Bureau of the Census, "Housing Starts", Construction Report Series C-22.
18- USDC, Bureau of the Census, U.S. Exports, Schedule E Commodity by Country. FT-410. [monthly and annual supplements].
19- USDA, Forest Service, "Volume and Value of sawtimber Stumpage Sold From National Forests by Selected Species and Regions". [quarterly and annual].
APPENDIX B
TABLE XVIII
AVERAGE HOURLY EARNINGS IN MANUFACTURING AND LUMBER & WOOD PRODUCTS (Current U.S. $)
Manufacturing SIC 24 ---------------------------- ----------------------------
Year US NW SO SW IL US NW SO SW IL
1950 1. 42 1. 76 1.18 1. 65 1.57 1.35 1.67 1.12 1.57 1.49 1951 1.53 1.91 1. 27 1.77 1.68 1.47 1.83 1. 22 1. 70 1.61 1952 1.67 2.01 1.33 1.87 1. 77 1.55 1. 87 1. 23 1. 74 1. 64 1953 1.77 2.07 1.40 1.97 1.86 1.62 1.89 1.29 1.80 1. 70 1954 1.81 2.13 1.45 2.03 1.91 1.63 1.91 1. 30 1.83 1.72 1955 1.89 2.22 1.50 2.11 1.99 1.68 1.97 1. 34 1.88 1. 76 1956 1.98 2.29 1.62 2.22 2.10 1. 76 2.04 1. 44 1.97 1.87 1957 2.07 2.34 1. 71 2.32 2.17 1.81 2.04 1. 50 2.03 1.90 1958 2.13 2.42 1. 76 2.44 2.24 1.89 2.15 1. 56 2.17 J .98 1959 2.22 2.52 1.82 2.53 2.33 1.97 2.23 1. 62 2.25 2.07 1960 2.29 2.59 1.89 2.62 2.40 2.07 2.34 1. 70 2.37 2.17 1961 2.35 2.66 1.93 2.72 2.46 2.11 2.38 1. 73 2.44 2.21 1962 2.38 2.74 2.02 2.80 2.54 2.14 2.46 1.81 2.52 2.28 1963 2.46 2.76 2.05 2.86 2.62 2.15 2.41 1. 79 2.50 2.29 1964 2.53 2.91 2.14 2.94 2.67 2.16 2.48 1.83 2.51 2.28 1965 2.61 3.00 2.20 3.02 2.75 2.18 2.51 1.84 2.52 2.30 1966 2.70 3.14 2.30 3.12 2.82 2.26 2.63 1.92 2.61 2.36 1967 2.82 3.27 2.39 3.25 2 :96 2.38 2.76 2.02 2.74 2.50 1968 2.99 3.42 2.56 3.40 3.10 2.53 2.89 2.16 2.88 2.62 1969 3.18 3.65 2.72 3.57 3.23 2.72 3.12 2.32 3.05 2.76 1970 3.36 3.98 2.91 3.81 3.48 2.98 3.53 2.58 3.38 3.08 1971 3.55 4.26 3.07 4.06 3.68 3.12 3.74 2.70 3.57 3.23 1972 3.77 4.42 3.14 4.23 3.91 3.25 3.81 2.71 3.65 3.37 1973 4.04 4.75 3.47 4.45 4.13 3.61 4.24 3.10 3.98 3.69 1974 4.37 4.99 3.70 4.59 4.45 3.87 4.42 3.28 4.06 3.94 1975 4.78 5.52 4.10 5.10 4.88 4.25 4.90 3.65 4.53 4.33 1976 5.15 6.01 4.43 5.46 5.30 4.66 5.43 4.01 4.94 4.80 1977 5.07 6.53 4.80 5.87 5.76 5.60 7.21 5.30 6.48 6.36 1978 5.61 7.53 5.39 6.47 6.77 6.07 8.15 5.84 7.00 7.33 1979 6.67 8.19 5.84 7.10 7.04 6.15 7.55 5.38 6.55 6.49 1980 7.18 8.85 6.35 7.60 7.68 6.55 8.07 5.79 6.93 7.01 1981 7.97 9.84 7.02 8.41 8.30 7.09 8.75 6.24 7.48 7.38 1982 8.50 10.69 7.65 9.25 9.01 7.54 9.48 6.78 8.21 7.99 1983 8.79 10.89 7.99 9.51 9.42 7.84 9.71 7.13 8.48 8.40 1984 9.14 11.00 8.28 9.75 9.66 8.02 9.65 7.27 8.56 8.47 ----------------------------------------------------------------Source: Mfg, Appendix A, (9] ; Lumber & Wood Prod., 1972-84, App.
A, (12], 1950-71 see page 80 of text for derivation.
APPENDIX C
TABLE XIX
INDEX OF LABOR PRODUCTIVITY IN LUMBER AND PLYWOOD PRODUCTION (1967=100)
Lumber Plywood --------------------------------------- ------------------Year US NW SO SW IL US NW SO
1950 0.46 0.00 0.00 0.00 0.00 0.41 0.53 0.00 1951 0.43 0.47 0.46 0.48 0.43 0.40 0.55 0.00 1952 0.48 0.55 0.47 0.56 0.51 0.42 0.58 0.00 1953 0.50 0.64 0.44 0.65 0.58 0.44 0.61 0.00 1954 0.55 0.70 0.48 0.71 0.64 0.46 0.64 0.00 1955 0.53 0.76 0.49 0.77 0.69 0.47 0.66 0.00 1956 0.56 0.81 0.52 0.83 0.74 0.51 0.69 0.00 1957 0.58 0.85 0.54 0.86 0.77 0.55 0.72 0.00 1958 0.63 0.88 0.56 0.90 0.81 0.59 0.75 0.00 1959 0.68 1.01 0.78 1.03 0.92 0.63 0.77 0.00 1960 0.71 0.92 0.59 0.93 0.84 0.64 0.78 0.00 1961 0.76 0.95 0.59 0.97 0.87 0.71 0.82 0.00 1962 0.71 0.92 0.65 0.93 0.84 0.70 0.82 0.00 1963 0.86 0.99 0.77 1.00 0.90 0.78 0.87 0.00 1964 0.93 1.04 0.89 0.95 0.96 0.86 0.96 0.86 1965 0.93 1.06 0.98 0.96 0.94 0.91 0.99 0.91 1966 0.96 0.98 0.97 0.98 0.91 0.92 0.95 0.92 1967 1. 00 1.00 1.00 1.00 1.00 1.00 1.00 1.00
. 1968 1.06 1.14 1.05 1.11 0.97 1.07 1.01 1.07 1969 1.00 0.96 1.03 1.02 0.90 0.99 0.95 0.99 1970 1.03 1.01 1.02 1.10 0.97 1.09 1.06 1. 09 1971 1.15 1.06 1.20 1.11 1.02 1.22 1.12 1. 22 1972 1.17 1.03 1.17 1.12 0.97 1. 23 1.16 1. 23 1973 1.18 1.03 1. 08 1.07 0.98 1.17 1.11 1.17 1974 1.11 1.02 0.95 0.92 0.86 1.05 1.08 1. 27 1975 1.09 1.09 1. 25 1.01 1.09 1.30 1.29 1. 65 1976 1.19 1.12 0.87 1.06 1.04 1. 34 1.41 1.94 1977 1.18 1.15 1.17 1.04 0.97 1.49 1.34 2.14 1978 1.16 1.12 1.02 0.99 0.96 1. 52 1.32 2.20 1979 1.13 1.05 1.11 1.00 0.98 1. 55 1. 25 2.26 1980 1.08 0.92 0.96 0.81 0.97 1.47 1.19 2.26 1981 1.05 0.97 1.42 0.81 1.05 1.56 1.34 3.04 1982 1.05 1.06 '1.30 0.85 1.16 1.57 1.33 3.31 1983 1. 32 1.34 1.43 1.06 1.19 1.60 1. 54 3.32 1984 1. 30 1.23 1.80 0.95 1.12 1.57 1.54 3.34 ------------------------------------------------------------Constructed from output/man-year estimates for each region. Source: See Appendix A, (1] , (5], and (6].
APPENDIX 0
TABLE XX
PRODUCER PRICE INDEX: TRANSPORTATION SERVICES AND LUMBER & WOOD PRODUCTS FREIGHT
Year
1950 1951 1952 1953 1954 1955 1956 1957 1958 1959 1960 1961 1962 1963 1964 1965 1966 1967 1968 1969 1970 1971 1972 1973 1974 1975 1976 1977 1978 1979 1980 1981 1982 1983 1984
Trans. Service 67=100
53.30 58.30 62.40 66.40 69.20 69.40 70.50 73.80 78.50 81.20 83.30 85.30 86.60 87.50 89.60 92.90 96.80
100.00 104.00 111. 30 123.10 133.00 136.00 136.90 141. 90 152.70 174.30 188.40 197.40 212.80 242.60 271. 60 294.40 303.60 313.80
Lum. & WP Freight 69=100
47.89 52.38 56.06 59.66 62.17 62.35 63.34 66.31 70.53 72.96 74.84 76.64 77.81 78.62 80.50 83.47 86.97 89.85 93.44
100.00 108.40 119.00 123.30 127.40 147.40 163.60 177.90 191. 70 205.70 233.80 273.90 320.50 350.60 355.00 382.50
-------------------------------------------------Source: Transportation Services, Appendix A, [14]. Lum & WP Freight, 1969-84, Append. A, [14]; 1950-68 derived from regression PPIWPF = -32.57 + 1.265 *(PPI Transp. Services).